AN ANALYSIS OF THE FLUTTER AND

DAMPING CHARACTERISTICS OF HELICOPTER ROTORS

A THESIS

Presented to

The Faculty of the Division of Graduate

Studies and Research

By

Sathy Padmanaban Viswanathan

In Partial Fulfillment

of the Requirements for the Degree

Doctor of Philosophy

in the School of Aerospace Engineering

Georgia Institute of Technology

January, 19 77
 AN ANALYSIS OF THE FLUTTER AND

DAMPING CHARACTERISTICS OF HELICOPTER ROTORS

Approved:

G. Alvin Pierce, Chairman

s~* "l i

Robin B. Gray ^

C. V i r g i T _ S m i t r r ^

Date approved by Chairman: J * 77?

i
 ii

ACKNOWLEDGMENTS

I wish to express my sincere gratitude to Dr. G. A. Pierce for his

kind guidance and assistance during this study. The many discussions 1

had with him greatly helped me understand aeroelasticity.

Grateful appreciation is extended to Dr. C. V. Smith for his

valuable suggestions for improvement. I would like to thank other members

of my reading committee, Dr. R. B. Gray, Dr. D. J. McGill, and Dr. Ueng

for their contributions.

I wish to thank Dr. M. B. Sledd for impressing upon me the

philosophical approach to Mathematical Physics. Dr. M. Stallybrass kindly

found the time to help me with problems in mathematics.

The discussions I had with Dr. V. R. Murthy greatly contributed to

my understanding of structural dynamic problems. Also I wish to thank

Dr. K. S. S. Nagaraja and Mr. R. Srinivasan for their helpful suggestions.

My sincere thanks to Mrs. Peggy Weldon for her patience and

skill in typing the thesis.

My parents and other members of my family made many sacrifices

during my academic career. My uncle the late Dr. S. Balakrishnan, my

aunt, and my grandmother greatly helped me during my school days, and

without their help, higher education may not have been possible for me.
 Ill

TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ii

LIST OF TABLES v

LIST OF ILLUSTRATIONS vi

NOMENCLATURE ix

SUMMARY xiv

Chapter

I. INTRODUCTION I

II. STRUCTURAL DYNAMICS OF A ROTATING BLADE 5

Equations of Motion and Boundary Conditions

Free Vibration Analysis
Example Blades
Orthogonality and the Generalized Equations
of Motion

III. UNSTEADY AERODYNAMICS AND FLUTTER EQUATIONS 42

Unsteady Rotor Flow Fieid

Loewy's Incompressible Aerodynamic Model
Two Compressible Aerodynamic Theories
Importance of Wake Effects
Derivation of the Flutter Equations

IV. The k-METHOD OF FLUTTER SOLUTION 63

Statement of the Problem

Determinant Method of Solution
The Conventional V-g Method or the k Method
An Example Problem
An Approximate True V-g Solution
V. THE p-k METHOD OF FLUTTER SOLUTION 96

The Concept of the Decay Rates

The Principle of the p-k Method
Substantiation of the p-k Method
 IV

TABLE OF CONTENTS (Continued)

Page

Two Numerical Schemes for the p-k Method

An Example Problem
A Brief Summary of the Various Methods

VI. UNSTEADY AERODYNAMICS OF THE p TYPE 122

The Mathematical Model

Governing Equations
Solution for the Pressure Distribution
Discussion of Results

VII. CONCLUSIONS AND RECOMMENDATIONS 139

Conclusions
Recommendations

APPENDIX

A. THE EIGENVALUE ROUTINE OF DESMARAIS AND BENNETT 142

REFERENCES 144
 LIST OF TABLES

Table Page
j.

1. M Matrix for Blade No. 1 at ft =12.53 40

rs
2. M Matrix for Blade No. 2 at 0. =12.53 41
rs
 VI

LIST OF ILLUSTRATIONS

Figure Page

1. Stability of One Aeroelastic Mode as a Function

of Rotor Speed 4

2. Blade Coordinate System 6

3. Variation of the Natural Frequencies of Blade

No. 1 (Hinged Root) with Rotor Speed 18

4. Variation of the Natural Frequencies of Blade

No. 2 (Cantilevered Root) with Rotor Speed 19

5a. The First Mode Shape of Blade No. 1 21

5b. The Second Mode Shape of Blade No. 1 *. . . . 22

5c. The Third Mode Shape of Blade No. 1 23

5d. The Fourth Mode Shape of Blade No. 1 at fi = 0 .... 24

5e. The Fourth Mode Shape of Blade No. 1 at fi =12.53 . . 25

5f. The Fifth Mode Shape of Blade No. 1 26

5g. The Sixth Mode Shape of Blade No. 1 27

5h. The Seventh Mode Shape of Blade No. 1 28

6a. The First Mode Shape of Blade No. 2 29

6b. The Second Mode Shape of Blade No. 2 30

6c. The Third Mode Shape of Blade No. 2 31

6d. The Fourth Mode Shape of Blade No. 2 32

6e. The Fifth Mode Shape of Blade No. 2 33

6f. The Sixth Mode Shape of Blade No. 2 34

6g. The Seventh Mode Shape of Blade No. 2 35

7. Schematic Elements of Unsteady Rotor Flow Field . . . . 43

 Vll

LIST OF ILLUSTRATIONS (Continued)

Figure Page

8. Schematic Representation of Unsteady Rotor Flow

Field 45

9. Loewy's Incompressible Aerodynamic Model 47

10. Compressible Aerodynamic Model of Hammond and

Pierce [13] for a Multibladed Rotor 49

11. Variation of the Modulus of Damping Ratio with

Frequency Ratio for a Pure Flapping Blade 53

12. Variation of the Phase Angle of the Aerodynamic

Moment with Frequency Ratio for a Pure
Flapping Blade 55

13. Positive Sign Convention for Unsteady

Aerodynamic Program ..... 59

14. Plot of the Flutter Determinant on the

Argand Diagram 65

15. Variation of Inflow Ratio with Blade Radius 76

16a. Frequency-Rotor Speed Plot of the First Mode 78

16b. Damping-Rotor Speed Plot of the First Mode 79

17a. Frequency-Rotor Speed Plot of the Second Mode 80

17b. Damping-Rotor Speed Plot of the Second Mode 81

18a. Frequency-Rotor Speed Plot of the Third Mode 82

18b. Damping-Rotor Speed Plot of the Third Mode 83

19a. Frequency-Rotor Speed Plot of the Fourth Mode 84

19b. Damping-Rotor Speed Plot of the Fourth Mode 85

20a. Bending Deformation of the Fluttering Blade No. 1 . . . 87

20b. Torsional Deformation of the Fluttering Blade No. 1 . . 88

21. Variation of Flutter Speed with Chordwise

Center of Gravity Location 90
 Vlll

LIST OF ILLUSTRATIONS (Continued)

Figure Page

22a. Frequency-Rotor Speed Plot of the First Mode 109

22b. Damping-Rotor Speed P l o t of t h e F i r s t Mode 110

23a. Frequency-Rotor Speed Plot of the Second Mode Ill

23b. Damping-Rotor Speed Plot of the Second Mode 112

24a. Frequency-Rotor Speed Plot of the Third Mode 113

24b. Damping-Rotor Speed Plot of the Third Mode 114

25a. Frequency-Rotor Speed Plot of the Fourth Mode 116

25b. Damping-Rotor Speed Plot of the Fourth Mode 117

26. p-Type Aerodynamic Mathematical Model for a Single

Bladed Rotor 123

27. Variation of L with Airfoil Motion Decay Factor . . . . 134

28. Variation of L with Airfoil Motion Decay Factor . . . . 135

ap
29. Variation of M, with Airfoil Motion Decay Factor . . . . 136
J
np
30. Variation of M with Airfoil Motion Decay Factor . . . . 137
ap
 IX

NOMENCLATURE

[A] generalized aerodynamic force coefficient matrix

A element of matrix [A] defined by Equation (45)

A undetermined coefficients of the various pressure modes

a nondimensional chordwise location of the elastic axis behind

the midchord

a two dimensional lift curve-slope

a free stream speed of sound

B undetermined coefficients of the various pressure modes

X/

b semi-chord of the airfoil

b c reference semi-chord
ref
E length of the vortex sheet considered in the analysis on
either side of the airfoil

EI..,EI bending rigidities about the major and minor neutral axes

EI bending rigidity in the out-of-plane direction

e distance by which mass axis lies ahead of the elastic axis

F vertical force per unit length of the beam

f time dependent part of the vertical force intensity

f time independent part of the vertical force intensity

f ,f abbreviations defined by Equation (12)

G wake integral function defined by Equation (9 7)

GJ torsional rigidity of the beam

GJ effective torsional rigidity defined by Equation (3)

g additional structural damping of the k method

g structural damping coefficient of the r-th vacuum normal mode

g. estimate of the damping present in the i-th aeroelastic mode
corresponding to the j-th scanning trial

h nondimensional wake spacing, see Figure 9

h' wake spacing, h' = hb

i imaginary number, /-l ; also an index for the modes

k reduced frequency, wb/^y

k polar radius of gyration of cross sectional area effective

in carrying tensile stresses, about elastic axis

k ..,k n mass radii of gyration about major neutral axis and about
ml mz . , . -, 1 , , ,
an axis perpendicular to chord through the elastic axis,
respectively
k polar radius of gyration of cross sectional mass about
m 2 2 2
elastic axis, k = k - + kn
m ml m2
L lift per unit span

L amplitude of the simple harmonic lift per unit span

L ,L lift coefficients defined by Equation (41)

L^,L lift coefficients defined by Equation (67)

L ,L nondimensional unsteady aerodynamic lift coefficients

defined by Equation (42)

L,
hp ,Lap p-type
r yr aerodynamic
y lift coefficients defined by jEquation
n v .(69)

M aerodynamic moment about elastic axis per unit span

63.
M amplitude of the aerodynamic moment about elastic axis per
unit span

M generalized mass of the r-th mode

M nondimensional generalized mass defined by Equation (49)

M element of the generalized mass matrix defined by Equation (26)

rs °
jyL ,M nondimensional unsteady aerodynamic moment coefficients
defined by Equation (42)

K ,M p-type aerodynamic moment coefficients defined by Equation (69)

 XI

M ,M moment coefficients defined by Equation (41)

x Z

M„,M moment coefficients defined by Equation (68)

M free stream Mach number, fty/a

7 J
00 ' 00

m frequency ratio, u)/Q

mf frequency ratio at flutter, a)f/fi_

m mass per unit length of the beam

m r reference value of m
ref
N number of vacuum normal modes considered in the flutter
analysis

n wake index number

n.. finite number of lower lying wakes considered in the p-type

aerodynamic model

p complex number denoting the decay rate and the frequency

of the exponentially damped simple harmonic motion,
p = yd) + ioj

p nondimensional value of p, p = p/a>

Q total torque about elastic axis per unit length of the

beam, or the number of blades in the rotor

q time dependent part of the torque per unit length

q time independent part of the torque per unit length

R radius of the rotor

R _ 2
T tension in the beam, J m r Q, dr
y
t time coordinate
V vertical climb rate of the rotor
00

v induced velocity on the airfoil

a
W total out-of-plane deflection of the beam

w time dependent part of the out-of-plane deflection

w time independent part of the out-of-plane deflection

 Xll

w amplitude of t h e out-of-plane deflection

w o u t - o f - p l a n e m o d e shape of t h e r-th v a c u u m n o r m a l m o d e
*
w n o n d i m e n s i o n a l o u t - o f -rp l a n e mode s h aKp e , w /b ,.
r ' r ref
wf nondimensional out-of-plane flutter mode shape

x,y,z coordinate system as shown in Figure 2

x' airfoil chordwise coordinate shown in Figure 26

A total torsional deformation of the beam

a time dependent part of the torsional deformation

a time independent part of the torsional deformation

a amplitude of the torsional deformation

a torsional modal deflection of the r-th vacuum normal mode

r
* *
a nondimensional torsional modal deflection, a = a
r ' r r
a (0) angle of incidence at the root

3 angle of incidence, or the nondimensional semi-chord, b/b f

A determinant of a m a t r i x

z distance of elastic axis behind t h e quarter chord axis in

terms of s e m i c h o r d s , e = 1/2 + a

r airfoil bound circulation

a
Y airfoil m o t i o n decay factor o r t h e strength of v o r t i c i t y o n
the airfoil and in the w a k e s

X. t h e i-th complex eigenvalue corresponding to the j-th scanning

I trial m . i n the k method
3
M v o r t e x v i s c o u s dissipation factor

ft angular velocity of t h e rotor

ft n o n d i m e n s i o n a l angular velocity, ft/w f

ftf angular velocity of the rotor at flutter

co frequency of vibration
 Xlll

LO frequency of vibration of the r-th vacuum normal mode

cof flutter frequency of vibration

2 ,— 4
<jj r- reference frequency, co _ = EI/ mR
n J
ref ref
to. estimate of the damped natural frequency in the i-th aero-
1
elastic mode corresponding to the j-th scanning trial

density of the free stream fluid medium

generalized force of the r-th mode

amplitude of the generalized force of the r-th mode

r
£ generalized coordinate of the r-th mode

£, amplitude of the generalized coordinate of the r-th mode

r
{£} column of the amplitudes of the generalized coordinates

5' airfoil chordwise coordinate

n nondimensional spanwise coordinate, n = y/R

Superscripts

differentiation with respect to spanwise coordinate

differentiation with respect to time coordinate

Prefixes

Re real part of a complex number

Im imaginary part of a complex number

 XIV

SUMMARY

Two relatively new methods of vibrational analysis of nonuniform

rotor blades in combined flapwise bending and torsion are reviewed.

The structural dynamic characteristics of an example blade are evalu-

ated using the Transmission matrix method and are later used in flutter

analyses.

An automated procedure is developed to obtain the matched flutter

point of a rotor in an axial flight condition. The determinant method

of flutter prediction turns out to be impracticable. By developing a

method called the approximate true V- g method, it is shown that the

failure of the k method to accurately predict the damping at subcritical

speeds is mostly due to the method of numerical solution.

The principles of the p - k method are explained and it is shown

that this method is well suited to analyze the damping and flutter char-

acteristics of rotor blades. An alternative numerical method of solution

is provided based on an eigenvalue analysis. An example flutter prob-

lem is solved by various methods. An unsteady rotor aerodynamic theory

of the p type is derived and the results from this analysis tend to

show that the implied assumption of the p-k method is sound.

 1

CHAPTER I

INTRODUCTION

All rotary wing aircraft such as helicopters, autogyros, VTOL

and STOL aircraft fitted with prop-rotors, are subject to the poten-

tially catastrophic phenomenon of rotor blade flutter. This aeroelastic

instability is characterized by self excited undamped oscillations of the

blade lifting surface in torsion and bending (elastic flapping). This

problem is generally solved by mass balancing the blade about the quar-

ter chord and designing the elastic axis to lie at the quarter chord

position. This solution usually results in added blade weight. Conse-

quently, the rotor hub has to be designed heavier to withstand the

increased centrifugal tensile forces.

Contemporary high performance main rotor systems are made of

light weight composite construction. The outboard sections of the blade

operate in the compressible subsonic Mach number regime, and are made

of cambered airfoil sections to improve the hover aerodynamic efficiency.

Furthermore, to augment the stability of the rotor in ground resonance,

air resonance, rotor-pylon aeromechanical stability, etc., the kinematic

aerodynamic coupling like flap-lag coupling, pitch-lag coupling, pitch-

flap coupling etc., are built into the system. These considerations

render the advanced rotor systems liable to a variety of potentially

dangerous aeroelastic instabilities, one of which is rotor blade flutter.

In the next decade, the rotor system designer will have a great need for
 2

being able to accurately predict the amount of stability present in

the various aeroelastic modes at subcritical speeds. These data are

very important for correlating with and guiding the flight flutter test-

ing and non-destructive wind tunnel testing.

In the last two decades, considerable work has been done in the

areas of rotor blade structural dynamics, rotor unsteady aerodynamics

and rotor aeroelasticity. The state of the art can now be considered

satisfactory in the area of structural dynamics. The task of obtaining

the unsteady aerodynamic forces on the helicopter rotor blade in forward

flight still remains formidable. The unsteady aerodynamic problem of a

rotor in hover or ascending vertical flight, or that of a prop-rotor in

the propeller mode of operation, seems to have been relatively well

solved. This thesis deals with the aeroelastic analysis of the rotor

under such an axial flight condition.

The structural dynamic principles are briefly discussed and an

example problem is solved in Chapter II. The aerodynamic theories are

reviewed and flutter equations of motion are derived in Chapter III.

The conventional method of flutter analysis consists of employing

an unsteady aerodynamic theory suitable for simple harmonic motion of

the lifting surface. By some approximate considerations, this method

provides an estimate of the stability present in the system at subcritical

speeds. This method is called the k method or the conventional V - g

method. While this method is satisfactory for prediction of the flutter

boundary flight condition, the estimation of the stability present in

the system is not acceptable at speeds below the critical speed. The k

method needs to be considerably modified before it can be employed for

 3

subcritical damping predictions. Results of one such analysis carried

out by Pierce and White [1] are shown in Figure 1. The modal damping is

oscillatory with respect to rotor speed and is even multivalued. They

recommended that the flutter criteria be based on the curve labeled

effective damping.

One main objective of this thesis is to explore methods that will

estimate the stability at subcritical speeds more accurately than the

results of Pierce and White. A relatively new method called the p-k

method has been highlighted by Hassig [2] in his study to improve the

damping prediction of fixed wings.

An aerodynamic theory is considered to be of the p type if it

deals with the motion of the lifting surface that decays in an exponen-

tially damped simple harmonic fashion. In general, all sophisticated

p type aerodynamic theories require excessive computer time. Hence the

p method of aeroelastic solution, which can be considered exact, is

numerically time-consuming.

However, if a k type (undamped simple harmonic motion) aerodynamic

theory is applied after suitable modifications to a p type motion, a

reasonably accurate and simplified formulation results. This is called

the p-k method. In Chapter IV the k method is discussed and in Chapter

V the p - k method is analyzed. Chapter VI contains a derivation of a

p-type rotor aerodynamic theory in an attempt to investigate the implied

assumption of the p - k method.

 Rotor speed
1 1 l 1 /
1 . 1 J
1 1 1 1 ' 1
f //I
Experimental Flutter / /
/ /
Speed / /

' Indicated
s/ Flutter
s J Speed
S 1
** j

Modal
Damping

1
fl \
Y-

J
1 j
[— Effective Damping

\ ,
k Method

V
Figure I. Stability of One Aeroelastic Mode as a
Function of Rotor Speed.
 5

CHAPTER II

STRUCTURAL DYNAMICS OF A ROTATING BLADE

The geometry of the rotor blade considered in this thesis is

shown in Figure 2. It possesses a smooth planform% and the local cen-

ter of gravity location gradually changes with the spanwise coordinate

y. The blade consists of symmetric airfoil sections with varying angles

of incidence relative to the x-y reference plane. The spanwise varia-

tion is constituted by the built-in twist (required to optimize the steady

aerodynamic performance of the rotor) and the aeroelastic twist (due

to the noncoincidence of the center of pressure axis and the elastic

axis). This variation in angle of incidence must be added to the angle

of incidence at the root (collective pitch) to obtain the local pitch

angle. The elastic axis is assumed to be straight.

Throughout this thesis, only torsional and out-of-plane (flapwise-

bending) deformations are considered. The edgewise (lead-lag) bending

deflections, due to vibration in the horizontal plane of rotation are

not considered. For the low inflow case considered here, the edgewise

bending oscillations are assumed not to produce any significant unsteady

aerodynamic forces.

A self excited vibrational phenomenon known as "ground resonance,"

which can be catastropic, has been experienced by several helicopters

and autogyros [3]. This phenomenon occurs frequently when the heli-

copter is supported on the ground by relatively soft tires, resulting in

 Straight
elastic axis

blade trailing edge

center of gravity axis

Figure 2. Blade Coordinate Svstem

 7

a low natural frequency of the machine in the sideward motion. The

resonance is characterized by the blade lagging vibrations coupled with

the vibration of the aircraft fore and aft, and sidewards. While ana-

lyzing self excited vibrations of this kind, blade lag vibrations are of

prime importance.

The main objective of this investigation is to develop some numeri-

cal programming schemes to automatically determine the true flutter

point. Edgewise oscillations are not considered throughout this thesis.

Another separate investigation must show the degree of validity of the

assumption of ignoring the in-plane oscillations. It is hoped that the

new principles brought about in this thesis regarding damping and flutter

analysis could be used for solving aeroelastic problems of similar

formulation.

Equations of Motion and Boundary Conditions

Houbolt and Brooks [4] have derived a comprehensive set of differ-

ential equations of motion for combined flapwise bending, edgewise bend-

ing and torsion of a twisted nonuniform rotor blade. The development

is based on the principles of beam theory, and secondary effects such as

deformation due to shear are not included. Other than assuming that the

elastic axis is straight, there are no major restrictions in their

derivation. The following additional assumptions are made here:

1) The distance between the elastic axis and the axis about

which the blade is rotating is zero at the root.

2) The distance between the area centroid of the tensile member

and the elastic axis is zero.

 3) The blade is untwisted and the mean angle of incidence at

every spanwise station equals the collective pitch.

4) In-plane deformation (edgewise bending deflection) is zero.

The first three assumptions are made simply because the computer

program available to carry out the vibration analysis does not have a

provision to include these terms. The flutter program to be developed

incorporates normal modes as input, so, if normal modes which include

these terms are available they can be used in an identical fashion pro-

vided the resulting governing equations are formally the same. The

last assumption has been discussed already.

With these assumptions, the governing differential equations

become:

-[(GJ )A']' " ft2myeW' + ft2m(k2 - k 2 )A

m mZ ml

_ 2 ••
+ m km A - m e W = Q (1)

-[EIWM]" + (TW')' - ( A y e A ) '

+ m(-W + eA) = -F (2)

where

GJ = GJ + Tk 2 (3)
m A

and

2 2
EI = EI cos 6 + EI sin 3 (A)
 9

Separating the time dependent and time independent parts of the external

forces and the resulting displacements

Q(y,t) = qQ(y) + q(y,t)

F(y,t) = f (y) + f(y,t)

o
(5)
W(y,t) = WQ(y) + w(y,t)

A(y,t) = aQ(y) + ot(y,t)

Then the following pairs of differential equations are obtained

-(GJ a ' ) ' - A y Je w ' + ft2m(k2 - k 2 J a = nq

mo o m2 ml o o

(6)
- ( E I w " ) M + ( T w ' ) ' - (ft 2 myea
J ) ' = -f
o o o o

(GJ a ' ) ' - A yJ ew" + ft2m(k2 - k 2 )a

m m2 ml

+ mka-mew =q
m ^
(7)

(EIw")" + (Tw1)' - ( A y e a ) '

+ m(-w + e a ) = - f .

In Equation (6) q and f contain time independent terms propor-

2
t i o n a l t o ft a s w e l l a s t h e mean o p e r a t i n g c o n s t a n t a e r o d y n a m i c forces.
 10

To obtain the mean aerodynamic forces knowing the deformation of the

blade in addition to the built-in twist, a theory like simple momentum

and blade element analysis,can be used. A more sophisticated theory like

vortex analysis or even good experimental results can be utilized. In

general, the relationship of the aerodynamic forces in axial flight with

respect to angle of attack at root is nonlinear. Equation (6) repre-

sents the static aeroelastic problem where the static equation of equili-

brium and the steady aerodynamic relationship must be simultaneously

satisfied. Nagaraja [5] has numerically obtained the solution for a

typical problem. The static aeroelastic solution establishes the mean

inflow; the wake spacing in the axial direction is then known. This

factor is important in evaluating the unsteady aerodynamic forces.

Equation (7) represents the dynamic equations of motion of the

blade. q and f are the resulting unsteady aerodynamic forces. A linear

aerodynamic theory would be employed to relate q and f to a and w.

Hence Equation (7) is linear and homogeneous. The following boundary

conditions would be employed in the solution.

For hinged root: w(y,t)| _ = 0

w"(y,t)|y=0 = 0 (8a)

«(y,t)|y = 0 = o

For fixed root

 11

w =
(y^)|y =0 °

w'(y,t)| Q = o (8b)

«(y,t)| = o

For free tip:

w"(y,t)|y=R = 0

[(EIw")' + Q2m Rea]| _ „ = 0 (8c)

y -K

«'(y.t)| R - o

The above are linear and homogeneous boundary conditions, satis-

fied by time dependent as well as time independent parts of the deforma-

tions A and W.
2
The Tk term contained in the (GJ ) term and the (T w1 )' term of
a m
Equation (7) show the effect of centrifugal forces in increasing the

effective torsional and bending stiffness of the beam. There are other

terms which arise because of elastic coupling and inertia loading due to

vibratory and centrifugal accelerations. The derivation is explained

in detail by Houbolt and Brooks [4]. For a nonuniform beam such as the

one considered here, GJ, k , m, e, k 0 , k , k , EI_, EI will be func-

a mz ml m 1 Z
tions of the spanwise coordinate y.

Free Vibration Analysis

In Equation (7), let

 12

a = a(y,t) = a(y) exp(iu)t)

w = w(y,t) = w(y) exp(iuot) (9)

f = q = 0

The following two coupled, homogeneous, ordinary differential equations

are obtained:

-(GJ a')' - f.w1 + f_a - mco2(-ew + k2a) = 0 (10)

m 1 2 m

-(EIw")" + (Tw,)f - (f^a)' - moo2(-w + ea) = 0 (11)

where

?1 = A y e (12a)

f0 = ^2m(k2„ - k 2 J (12b)
2 mz ml

The boundary conditions that a and w should satisfy are given by

Equation (8) by replacing w and a by w and a.

Thus a and w satisfy homogeneous differential equations with homo-

geneous boundary conditions. If there exists an w, say w , correspond-

ing to which a nontrivial solution exists, then a natural frequency, oo ,

and a vacuum mode shape, a = a and w = w, are obtained. Some numeri-

cal techniques to solve this problem are discussed in detail by Murthy

[6]. Two of the methods are briefly summarized here in the interest of

completeness.
 13

Transmission Matrix Method

Let {Z(y)}be a column vector defining the states at the spanwise

station y, given by

w(y)

w'(y)

ot(y)
(Z(y)} = •< (13)
Qy(y)

M (y)
x
V (y)
Z
J

The governing linear differential equations of vibration can be written

as a set of first order equations in matrix form as

(Z(y)} = [A(y)]{Z(y)} (14)

dy

The transmission matrix [T(y)J is defined by

(Z(y)} = [T(y)]{Z(0)} (15)

I t can be shown t h a t

f- [T(y)] = [ A ( y ) ] [ T ( y ) ] (16)

By shrinking y to 0, it is noted that [T(0)] is an identity matrix.

From Equation (15)

(Z(R)} = [T(R)]{Z(0)} (17)

 14

Applying the boundary conditions at the root and the tip regarding

bending deflection, slope of neutral axis, torsional deflection, shear

force, bending moment, and torque, part of Equation (17) can be written

as

{0} = [T1(R)]{Z1(0)> (18)

where [T (R)] is a partitioned matrix of [T(R)], and {Z (0)} is the

corresponding part of {Z(0)} representing the nonvanishing quantities

at the root. Clearly, for a nontrivial solution for {Z (0)} to exist,

|[T1(R)]| = 0 (19)

which becomes the characteristic equation. The elements of [T (R)] are

obtained through Runge-Kutta numerical integration. S. Rubin [7] is

one of the pioneering investigators of this method and Murthy [6] has

extended it to cover the vibrational analysis of a very general case of

a rotating blade.

Several trial frequencies are chosen in an increasing sequence.

The frequency determinant is evaluated at each trial argument and the

vanishing of this determinant corresponds to a natural frequency. This

frequency choice method has one disadvantage in that if the determinant

function is not carefully analyzed, two or more of its zeros may go

undetected. Hence caution must be exercised when two natural frequencies

are expected to be close together.

Using the above obtained natural frequencies, the boundary condi-

tions of the problem, the transmission matrix obtained through integration,

 15

and the information contained in Equation (18), the mode shapes are

readily obtained. One very outstanding feature of this elegant method

is that the mode shapes could be obtained at as many spanwise stations

as desired without increasing the order of the frequency determinant.

Of course, the numerical round off errors might grow if the Runge-Kutta

interval of integration is very much reduced.

Integrating Matrix Method

Let g(x) be a continuous and smooth function of one independent

variable x in the interval x < x < x . Let the interval be divided

o — — n

into n equal subintervals and let the values of g be known at these

(n+ 1) interval points as g(x.), j = 0,1,2,...,n. Assume that g(x) can

be represented approximately by a polynomial of degree r(r <^ n ) . This

polynomial equation may be expressed in the form of Newton's forward-

difference interpolation formula. Using this formula the function can

be integrated analytically in terms of the values of the function at the

end points of the sub-intervals. The following matrix equation can then

be written:

j g(x)dx

x„
(20)
/ g(x)dx - ifi y
X

/ g(x)dx
n
 16

The matrix [^] of Equation (20) is called the integrating matrix.

Premultiplying the column matrix consisting of the values of the func-

tion at the chosen points by [j], the integrals of the function are

obtained.

The solution to the governing differential equations of motion

is developed entirely in matrix notation which allows the numerical

solution to be developed in a compact and orderly fashion. The matrix

differential equations are then integrated repeatedly by using the

integrating matrix as an operator. Next, the constants of integration

are evaluated by applying the boundary conditions. Finally, the result-

ing matrix equation is expressed in the familiar concise form of the eigen-

value problem. Hunter [8] used this method to study the vibrational

characteristics of propeller blades.

An outstanding feature of this method is that, when carried out

numerically accurately, the frequencies of vibration are obtained

rapidly. Such an initial estimate could profitably be used as the

input to a more sophisticated method like the transmission matrix method

and thus the eigenvalues could rapidly be refined.

Example Blades

Although several assumptions have been made and discussed, Equa-

tions (10) and (11) still represent a sophisticated description of the

problem. The transmission matrix approach is a powerful method to obtain

the normal modes accurately. Two example blades have been chosen and

their mode shape and frequencies have been computed by the computer pro-

gram prepared by Murthy [6] which utilizes the transmission matrix. It

 17

is believed that the seven modes obtained and described below for each

blade represent an accurate description of their structural dynamic

properties. An attempt is made to provide all the details of the

blades and the results of the vibration analyses, since it will be a

useful reference in the literature. The blades chosen are nearly the

same as the model blades tested by Brooks and Baker [9]. It is believed

that the example blades will provide a realistic and informative pic-

ture of rotor aeroelasticity.

The example blades numbered 1 and 2 are identical except that

Blade No. 1 is hinged at the root whereas Blade No. 2 is fixed at the

root. These two blades have the following uniform properties along the

span:

m = 0.00135 slug/inch

EI = 26000 lb-inch2

GJ = 10000 lb-inch2

R = 46.0 inch

b = 2.0 inch

e = -0.45 inch.

k _ = 0.1 inch
ml
k . = 0.976 inch
m2
kA = 0.948 inch.
A

The above given data are sufficient to determine the normal modes of

the blades.

Figures 3 and 4 show the variation of the natural frequencies of

Blades No. 1 and 2 respectively, with rotor speed. The strong effect

of centrifugal forces in stiffening the blade Is reflected in the

 18

800

600

L400

1200 -

CO
n

L000

rad
scc

800

600

400

200

30 90
m [rad/sec
_±J_ At IJ J50

Figure 3. Variation of the Natural Frequencies of Blade No

(Hinged Root) with Rotor Speed.
 19

1800

600

6Jn J 400
CrcioL/s)

J 200 L

JOOO h

800 U

600 r

400 h

2 00

30 60 90 f^ . . , , .1 50
JL2. [ rad/sec J

Figure 4. Variation of the Natural Frequencies of Blade No. 2

(Cantilevered Root) With Rotor Speed.
 20

monotohic increase of the natural frequencies with rotor speed. All the

points lying on a straight line though the origin in these two Figures,

are such that they have the same value for the ratio of natural frequency

to rotor speed. It is conventional to draw these fan lines because the

frequency ratio can then be readily seen at any point of interest.

Figures 5(a) through 5(h) illustrate the first seven mode shapes

of Blade No. 1 at ft = 0 and ft = 12.53. Figures 6(a) through 6(g)

illustrate the first seven mode shapes of Blade No. 2 at ft = 0 and

ft* = 12.53.

Figure 5(a) shows that the first mode of Blade No. 1 is essen-
JL

tially a flapping mode with no torsional deflection at ft =* 0 and very

small torsional deflection at ft =12.53. Figure 3 shows that the natu-

ral frequency of this mode is essentially the same as that of the rotor

speed. The graph of the first mode shape of Blade No. 2 can be seen

in Figure 6(a). The bending part is the first cantilever mode and there

is little torsional deflection. For ft > 30 rad/sec., the frequency of

this mode is only slightly higher than the rotor speed.

Figure 5(b) shows the second mode shape of Blade No. 1. This is

predominantly a bending mode at lower rotor speeds. Figure 3 shows the

considerable influence of rotor speed in increasing the natural fre-

quency of this mode. Figure 6(b) is the graph of the second mode shape

of Blade No. 2; this is also a predominantly bending mode at low rotor

speeds.

It is generally observed that modes exhibiting predominantly

torsional deflections are relatively unaffected by rotor speed in terms

of changes in natural frequency. Figures 3 and 4 show small increases

-0.8

0.4

- , _/2- - W , / i JO -

0.0 1 i * t

1 1 I 1 1 1 1 1 1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 yp 0.9
- -

-0.4

Figure 5a. The First Mode Shape of Blade No. 1

-0.8 U

0.4

\-

1 — , — .

0.0 1 —I 1 - \ ' - —i 1- 1
0.1 0.2 0.3 0.4 \ 0.5 0.6 0.7 0.9
u \—-dT=o
0.8
P —

-0.4

NJ
Figure 5b. The Second Mode Shape of Blade No. 1. NJ
0.8

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Figure 5c. The Third Mode Shape of Blade No. 1.

 0.4 _

*> > -
WfO?)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 ' 0.9
1 i 1 1 1 i i i
i •"' 1 " ' —.— • i i
0.0 I r 1 r

-0.4

-0.8 -

Figure 5d. The Fourth Mode Shape of Blade No. 1 at ft = 0. -o

-0.4
N3
Figure 5e. The Fourth Mode Shape of Blade No. 1 at 0 =12.53 Ul
 0.8

0.4

0.0 -

-0.4 -

-0.8

-0.4

Figure 5f. The Fifth Mode Shape of Blade No. 1 ro

 .4
- _ / 2 * = /2-53
wjc?) r— ^

0.0 "~ i 1 i • i T^-ta. • i •

1 1 ••
1 i 1 1--^—^ulL—T_ 1 \
0.1 0.2 0.3 0.4 0.5 0.6 ^ ^ ^ ^ 0.9
=^&4-. PJ
-0.4

Figure 5g. The Sixth Mode Shape of Blade No. 1

-0.8

-0.4 L

Figure 5h. The Seventh Mode Shape of Blade No. 1 KJ

CO
 0.8
W
,V
0.4

r i /

u . U

-0.4

-0.8

-0.4

Figure 6a. The First Mode Shape of Blade No. 2 ro

^o
 0.4

a*c7) -
rsi*-o
_Q*=/2-53-^ —

0.0 —1 —i 1 1 1
0.1 0.2 0.3
1 \
-
0.4 0.5 0.6 0.7 0.8 7 °-9
-0.4

Figure 6b. The Second Mode Shape of Blade No. 2 LJ

o
 0.4

w,*<7>
0.2 0.3
0.0

-0.4 Jl = 12-53

-0.

Figure 6c. The T h i r d Mode Shape of B l a d e No. 2 .

 0.8 -j
VJ^) •J

0.4 -J

0.0
0.1
[
0.2
1
0.3
i
0.4
1 W
j f 0.6
i
0.7
i
0.8
l
\V
\ \
0.9
1
p 1
1 1 1 \ JOT 1 1 1
W '
H

-0.4 ^ j f = 12-53—^

\ 1
^-Jl*=0
-0.8

0.4

-0.4

Figure 6d. The Fourth Mode Shape of Blade No. 2 OJ

ho
-0.8

-0.4

Figure 6e. The Fifth Mode Shape of Blade No. 2.

 0.8

ocfa) Jl^O -J

0.4

A
1 1 1 • I 1 JV 1 i i
0.0 1 1 t 1 1 Jr 1 1
0.1 0.2 0.3 0.4 0.5 0.6 / f 0.7 0.8 /o 0.9
A

-0.4 -J

-0.8 -J

Figure 6f. The Sixth Mode Shape of Blade No. 2. U:

-P-
 0.8

W 0?) -

0.4

0.0

-0.4

-0.8

-0.4

Figure 6g. The Seventh Mode Shape of Blade No. 2

 36

for the natural frequencies of the third mode whereas large changes are

observed for the fourth mode of both blades. At ft = 0, because of the

close vicinity of third and fourth modes, the third mode exhibits con-

siderable bending (see Figure 5(c)). But at ft* = 12.53, the bending

part of this mode loses both the nodal points and the mode becomes pre-

dominantly torsional. This demonstrates the complex interaction of

centrifugal forces and bending-torsion coupling effects. However, the

third mode shape of Blade No. 2 as can be seen plotted in Figure 6(c)

is predominantly torsional. There is also a considerable separation

in the natural frequencies of the third and fourth modes for this blade

even at low rotor speeds.

For Blade No. 1, the fourth mode also has a drastic change in the

mode shape as the rotor speed changes from ft = 0 to ft = 12.53 (see

Figure 5(d)). But Blade No. 2 (see Figure 6(d)) exhibits predominant

bending oscillations in the fourth mode. The natural frequencies of the

fourth mode for both blades increase considerably with rotor speed (Fig-

ures 3 and 4).

The fifth and seventh modes of both the blades are predominantly

bending and their natural frequencies are highly dependent on rotor

speed (see Figures 3,4,5(f),5(h),6(e) and 6(g)). It can also be noted

that the mode shapes of the fifth and seventh modes of one blade are

similar to those of the other blade. Comparing the plots in Figures 3

and 4, it is seen that the fixed root results in higher frequencies

than hinged root, in case of predominantly bending modes.

The sixth mode for either blade is predominantly torsional as

shown in Figures 5(g) and 6(f). As can be expected, the difference in

 37

the root boundary condition has not affected either the frequency or

the mode shape (see Figures 3 and 4). This is because of the fact that

as far as torsional deformation goes, hinged as well as fixed roots pro-

vide only a fixed boundary condition. The frequencies of the third mode

for both the blades are approximately equal at all rotor speeds; this

is another indication that this mode is also predominantly torsional.

For both the blades, the bending deflections of the fifth, sixth

and seventh modes show respectively 3, 1, and 4 nodal points. Although

generally an ordered increase in the number of nodal points can be

expected with increasing mode number, it is believed that this need not

always happen. The frequency determinant was examined carefully at

some high rotor speeds and the behaviour of the determinant function

was found to be satisfactory. It is further believed that the seven

modes shown in the graphs for each blade are free of significant numeri-

cal errors.

A final remark is made regarding the effect of rotor speed on mode

shapes. For the fifth, sixth and seventh modes of either blade, the

mode shapes are not very different for ft = 0 and ft = 12.53. For the

Blade No. 2 it can be said that for 0 _< ft <^ 12.53, all the mode shapes

are similar at least in "pattern," but for Blade No. 1, the third and

fourth modes are considerably different at different rotor speeds. These

factors must be considered if an attempt is made to use non-rotating

mode shapes in a dynamic response analysis of a rotating beam.

Orthogonality and the Generalized Equations of Motion

The mode shape of the r-th mode, a (y) and w (y), and the frequency
 38

(i) , satisfy the differential equations (10) and (11) and boundary con-

ditions (8). An orthogonality property can be derived which states

that for OJ ^ m , i.e., for the r-th and s-th modes possessing distinct
r s
frequencies,

_R _
i m{w w + k a a - e(w a + w a )}dy = 0 (21)
; v
r s m r s r s s r
o

If a solution is sought of the form

a(y,t) = I ar (y)£r(t)
r=l
(22)

W(y,t) = I w (y)£ (t)

r=l

for the differential equations (7), then the normal coordinates £ (t)

are obtained by solving

Mr£r(t) + o)^Mr?r(c) = H r (t), r = 1,2,3,... (23)

where

R
_ 2 2 2
M = [ m{w + k a - 2ew a >dy (24)
r ; r m r r r y v
'
o

and

R R
(t) = / f(y,t)w r (y)dy + / q(y,t)a r (y)dy (25)
 39

When f and q are known, 5 (t) can be computed and £ (t) can be deter-
r r

mined from Equation (23). Then the response of the beam is

obtained from the modal series of Equation (22).

For the uniform example blades, Equation (21) implies that

r * * 2 * *
J tw w + r a a (26)
J v
r s a r s '
o 0 if r 4 s
- Cr) [w a +w a ]}dn=M = M
b r s s r rs sr M
r
— 2 it r= s
mb R
M has been evaluated for Blades No. 1 and 2 at several rotor speeds.
rs
It has been observed to form an almost diagonal matrix at every rotor

speed. Tables 1 and 2 illustrate the 7 x 7 M matrix for both the

rs

blades at ft = 12.53. The nonvanishing of the off-diagonal elements is

due to the numerical inaccuracy of the normal modes shown in Figures

5(a) through 5(h) and 6(a) through 6(g). However, the results can be

considered quite satisfactory.

 Table 1. M Matrix for Blade No. 1 at Q =12.53
rs

\ s
r
\ 1 2 3 4 5 6 7

1 0.33333

2 -0.033xl0"6 0.19320

3 -0.051x10"6 -0.074xlO~6 0.096265 M =M

rs sr
6 6 6
4 -0.137xl0" 0.208xl0~ 0.341xl0~ 0.26371

5 -0.302xl0~6 -0.330xl0"6 0.357xl0~6 0.330xl0~6 0.25211

6 -0.062xl0~6 -0.098xl0"6 -0.010xl0"6 -0.327xl0"6 -1.223xl0"6 0.091576

7 -1.072xl0~6 -0.384xl0~6 0.351xl0~6 -1.652xl0"6 -2.394xl0~6 2.677xl0~6 0.27620

4>
O
 Table 2. M Matrix for Blade No. 2 at H =12.53
rs

1 2 3 4 5 6 7
\ -

1 0.29036

2 -0.248xl0"6 0.18344 M =M
rs sr
3 -0.017xl0~6 -0.228xl0~6 0.097163

4 0.059xl0" 6 -0.268xl0~6 0.389xl0~ 6 0.23654

5 -0.761xl0~6 0.483xl0~ 6 0.148xl0~ 6 -1.885xl0~6 0.26795

6 -0.162xl0~6 0.017xl0~ 6 -0.016xl0"6 -0.869xl0"6 -0.536xl0~ 6 0.092588

7 -0.697xl0"6 -1.714xl0~6 0.919xl0" 6 1.482xl0~ 6 -lO.OxlO" 6 2.l82xl0-6 0.30184

 42

CHAPTER III

UNSTEADY AERODYNAMICS AND FLUTTER EQUATIONS

In this chapter, the flow about a rotor operating at a steady

mean inflow with small simple harmonic perturbations in flow parameters

is briefly discussed, and some theories available to determine the

unsteady forces are reviewed. A simple example is used to illustrate the

importance of considering the wakes below the rotor. The flutter equa-

tions with this flow phenomenon are derived.

Unsteady Rotor Flow Field

A rotor in hovering or ascending vertical flight trails a tip

vortex which is blown axially downward so that, if otherwise undisturbed,

it would form a contracting helix as shown in Figure 7(a). For simpli-

city, consider that the inflow over the disc, u, is constant; then, the

fluid that comes off the trailing edge of the blades makes a helical

surface with horizontal radial elements (see Figure 7(b)).

Now, if there is an oscillation in blade effective angle of

attack, blade lift will alternate also, and as a result of these changes

in lift, vortices will be shed continuously at the blade trailing edge.

These vortices fall along the horizontal radial elements of the helical

surface shown in Figure 7(b), so long as the oscillations in angle of

attack are small. Figure 7(c) shows this helical sheet of shed vorticity.

Vorticity is considered to be on the helical surface and the vertical

displacements from that surface (as in Figure 7(c)) represent the strength
 43

CL.

igure 7. Schematic. Elements of Unsteady

Rotor Flow Field.
 44

of the vorticity at a particular azimuthal and radial position. The

variation in the vortex strength around the azimuth corresponds to the

history of the motion of a given blade element at a fixed radius; the

variation of the shed vortex strength in the radial direction at any

fixed azimuth angle is a function of the variation with the blade span

of 1) blade chord, 2) amplitude of the oscillation in effective angle

of attack and 3) relative air velocity. A variation of shed vorticity

in the radial direction implies the existence of trailing vortices at

constant radii similar to and inboard of the tip vortex. These trailing

vortices have been included in Figure 7(d).

The schematic drawings of Figure 7(a) through 7(d) indicate pic-

torially the complexities of attempting to obtain a complete represen-

tation of unsteady rotor aerodynamics. When 2^u/Q^, the vertical spacing

between adjacent helical surfaces of shed vorticity, is very large, then

one would expect that all shed vorticity beyond a small fraction of a

revolution would be too far below the blade in question to have an

effect on the blade loading. Under these conditions, it would be suffi-

cient to account for only the attached vortex sheet within that fraction

of a revolution, as in Figure 8(a). On the other hand, when 2TTU/Q^ is

very small, all the sheets of shed vorticity tend to pile up on each

other, and the effect of that vorticity close to the blade in question

(shed by the several previous blades and/or in the several previous revo-

lutions) is of more importance than that which exists beyond a small

aximuth angle on either side of the blade. This situation is depicted in

Figure 8(b).
 45

a. High Inflow b. Low Inflow

Figure 8. Schematic Representation of Unsteady

Rotor Flow Field.
 46

The first condition exists at high rotor thrust coefficients; the

second condition is associated with low thrust coefficients and is

encountered in "wake-flutter." Only the case of low inflow flutter is

considered in this thesis. Henceforth, the aerodynamic models that

are explained and used are for low inflow cases.

Loewy's Incompressible Aerodynamic Model

In arriving at a model which is mathematically tractable for the

case of low inflow, it is assumed that only the vorticity contained within

a small double azimuth angle straddling the blade is of real consequence.

The flow problem at a given blade radius is considered two dimensional and

this theory can then be applied in a strip theory fashion to a three-

dimensional rotor. The portion of the circular cylindrical surface which

is determined by 1) a particular blade radius, 2) the azimuth angle on

either side of the blade section (within which the shed vorticity is of

importance), and 3) the vertical distance spanned by a given number of

rows of vorticity, can be developed and projected on to a plane - one

in which the two-dimensional unsteady aerodynamic problem may be

attacked. The above considerations form the basis for the incompressible

flow model suggested by Loewy [10]. This model is shown in Figure 9.

The aerodynamic lift and moment acting on the airfoil are evaluated in

terms of nondimensional coefficients which are functions of reduced fre-

quency, frequency ratio, inflow ratio, number of blades and the phase

differences in the oscillation of other blades in the rotor with respect

to the reference blade. In the case of compressible theories, the Mach

number would also be included in the list of parameters on which the

aerodynamic coefficients depend.

 47

Reference Airfoi

0 2
GLhir
QhJr

0 Q-L

J 0

q = blade number

n = «° q = Q-. n = rotor revolution index

F i g u r e 9. boewy's I n c o m p r e s s i b l e Aerodynamic Model.

 48

Two Compressible Aerodynamic Theories

Jones and Rao [11] have solved the above problem for compressible

flow utilizing a model very similar to that of Loewy [10]. Their

analysis of the problem follows a technique developed earlier by Jones

[12] for a two-dimensional fixed wing oscillating in subsonic flow.

Coincident with the work of Jones and Rao, Hammond and Pierce [13]

independently analyzed a slightly different model of the two dimensional

compressible problem. Their model is illustrated in Figure 10. By intro-

ducing the acceleration potential, the governing integral equation for the

flow and its attendant downwash boundary condition are developed and solved

numerically using a pressure mode assumption and a collocation technique.

Hammond and Pierce [13] have shown that for small values of the frequency

ratios near and above 1, the aerodynamic coefficients are in good

agreement.

Pierce and White [1] employed the two compressible theories men-

tioned above, to predict the flutter speed of a model which had been

flutter tested by Brooks and Baker [9]. Both the theories predicted

flutter speeds which agreed well with the experimental results of

Brooks and Baker. Frequency ratio is a dominant factor in the flutter

analysis at low inflow. The flutter frequency ratio for the above case

was 2.3 and corresponding to this value, the two theories are in close

agreement. White [14] concluded from some theoretical flutter calcula-

tions that the theories of Jones and Rao [11] and Hammond and Pierce

[13] predict essentially the same flutter speeds but the former theory

requires significantly less computer time. For the flutter calculations

of this thesis, the theory of Jones and Rao [11] will be used.
 Reference Airfoil

v=Jir y-u •Jr

1
o o
•6,r zhb
C
e2 r k>
C

QhJr
Q.hir

•2rtr+otr

q = Blade number

n - °o Q- 1 n = Rotor revolution index

2 7,
q Q

Figure 10. Compressible Aerodynamic Mode] of Hammond and

Pierce |13| for a MuJtibladed Rotor.
-P-
vC
 50

Importance of Wake Effects

A simple example is given to show the importance of having to

take into account the velocity induced by the vortex sheets that lie

in the wake of the rotor.

Consider a single bladed rotor with uniform blade section proper-

ties along the span and with infinitely large bending and torsional

stiffness. Let it be provided with a flapping hinge at the root and let

6(t) denote the flapping response in radians at time t. The equation

of motion of this single degree of freedom system can be written as

i j + fi2I,3 = T(t) + / ~ rdr

< 27 >
J
ft dr
o

where

r ~ 2 . = m R /3 /OON
i. = J m r dr (28)

T(t) is the externally applied moment about the flapping hinge in same

sense as 3(t) and dL/dr is the lift per unit span at blade section r

caused by the flapping motion. Equation (27) can be viewed as the

balance of inertial moment, centrifugal restoring moment, externally

applied moment and aerodynamic damping moment about the flapping hinge,

Consistent with the theory of the vibration of single degree of

freedom damped systems, a damping ratio, £, can be defined through

R dL
/ ^ • rdr = - 2 I f n C 3 (29)
o
 51

Since the aerodynamic moment is known to be not completely in phase with

the velocity, a real number t, cannot satisfy the above equation. In

this example, a simple harmonic motion for flapping response is con-

sidered. Thus t, as defined by Equation (29) will turn out to be a time

invariant complex number denoting the phase and magnitude relationship

of 3 with respect to aerodynamic moment.

Corresponding to

3 = 3 Q exp(iojt) (30)

unsteady aerodynamic strip theory y i e l d s

2 2
dL/dr = TTp^co b L r 3 exp(io)t) (31)

where the aerodynamic coefficient L (r) is symbolically written as

Lh(r) = Lh(k(r),m,h(r),Mco(r)) (32)

It can now be shown that

1 2
z, = 1*5 im( / Lh(n)n dn)/v (33)
o

where
— 2
\i = d e n s i t y r a t i o = m/irp b

Since t h e reduced frequency and Mach number a t any s p e c i f i e d spanwise

 52

s t a t i o n a r e known from t h e frequency r a t i o , t h e blade, planform and Mach

number a t the t i p , i t can be w r i t t e n t h a t

•Q = C ( m , R / b , Moot± ,h,y) (34)

In this example, R/b = 23 , M _ . = 0 . 7 , h is constant over the

ootip '
rotor disc at 3, u = 80. C, then, is only a function of the frequency

ratio, this function being determined by the aerodynamic theory used.

In Figure 11, the modulus of Z, is shown plotted against frequency ratio,

according to four different aerodynamic theories.

Curve 1 was generated by employing the compressible theory of

Jones and Rao [11]. This is one of the most comprehensive theories

available at the moment. Curve 2 was generated by the same theory except

that M ^.
tip was deliberately
J
set equal
-± to zero,
> so the differences between

curves 1 and 2 can be considered to indicate the compressibility effects,

In the interest of clarity, curve 1 is not shown for m > 2.75, but the

general relationship between curves 1 and 2 remains the same for m > 2.75,

Curve 3 was obtained by employing fixed wing compressible, unsteady

aerodynamics [12] in a strip theory fashion. Curve 4 was obtained by

assuming fixed wing steady aerodynamic strip theory to this unsteady

case wherein the ratio of the velocity induced by flapping to the

equivalent forward speed (Qr) is considered to constitute the angle of

attack. Compressibility is accounted for by employing Prandtl-Glauert

correction. In this case, L is given by

U = -2i/(k/l-M
=-2i/(k/l-M ) (35)
0.25

0.20 -

C
0.15

0.10

Jones and Rao theory [11]

0.05 Loewy's theory [10|

Fixed wing compressible theory 12

Steady aerodynamic: theory

m
0.0
o.o L.O 2.0 3.0 0 5.0 6.0

Figure 11. Variation ol the Modulus of Damping Ratio with

Frequency Ratio for a Pure Flapping Blade.
 54

According to the. theory of Jones and Rao, which is believed to

be sufficiently accurate, there is a very significant drop in the aero-

dynamic damping moment at integer values of frequency ratio. At low

integral values of m, this drop is confined to a small neighborhood near

the integral values, but at higher values of m, this width of low damping

increases. In as far as this feature is concerned, curves 1 and 2 pre-

dict the same, i.e., compressibility effects do not change this behavior.

The striking difference between curves 1 and 2 is that curve 2 always

shows a lower aerodynamic moment. This is intuitively expectable since

for steady flow, the Prandtl-Glauert similarity rule predicts increased

lift at higher Mach numbers for the same angle of attack.

Although curve 3 shows values for £ in the same range as shown by

curves 1 and 2, the fixed wing theory completely fails to predict loss

of damping at integer values of m, because it ignores the helical vortex

surface lying in the wake of the rotor. As such, fixed wing theory should

be considered unsuitable for unsteady rotor flow problems at low inflow

conditions.

The steady theory predicts a constant value of |£| at 0.266. As

seen in Figure 11, this theory is apparently satisfactory for m £ 0.75

in this example. It is interesting to note that the model chosen in

this example shows considerable damping at low frequency of flapping.

Figure 12 shows the phase angle, <f>, by which the aerodynamic

R
dL
moment, / -r- r dr, leads the flapping deflection, $. Compressible
o
as well as incompressible rotor aerodynamic theories predict that at
integer values of frequency ratio, the phase angle is close to -90°,
 60"

-90°

9
-120°

Jones and Rao theory [11]

Loewy's theory [10]

-150° Fixed wing compressible theory [12

Steady aerodynamic theory

! «n°
0.0 1.0 2.0 .0 4.0 5.0 6.0 7.0
m
Figure 12. Variation of the Phase Angle of the Aerodynamic Moment
with Frequency Ratio for a Pure Flapping Blade.
 56

i.e., the damping moment is approximately 180° out of phase with velocity

of flapping. However, away from integral values of m, the phase angle

differs considerably from -90°. Again fixed wing unsteady theory and

steady theory do not predict such an oscillatory behavior for the phase

angle. Though not shown in Figure 12, curves 1 and 2 exhibit the same

kind of oscillatory behavior for higher values of m.

White [14] conducted flutter analyses of a model rotor by employ-

ing the compressible theory of Jones and Rao [11] once with the wake

terms included and another time without including the wake terms. The

latter case yielded a flutter speed which was considerably higher than

the reported experimental result of Brooks and Baker [9]. This case was

for a blade pitch angle of 7.2° and lower pitch angles may produce even

larger errors when the wake is neglected.

Experimental results of Ham, Moser and Zvara [15] and Daughaday,

Du Waldt and Gates [16] also show considerably decreased aerodynamic

damping at integer values of the frequency ratio.

Derivation of the Flutter Equations

Flutter is the phenomenon where the lifting surface undergoes

undamped simple harmonic oscillations without the application of any

external forces. At this frequency of vibration, the elastic restoring

forces, the inertial forces, and the internal structural damping forces

are in equilibrium with the aerodynamic forces which are created solely

because of the oscillation of the surface. Flutter is a stability boundary,

The subsequent response of the aeroelastic system to a disturbance either

decays or grows depending upon whether the speed is below or above the
 57

flutter boundary. The problem is treated as linear and homogeneous; the

amplitude of vibration is arbitrary.

The expression "flutter mode" means the mode of oscillation in

which flutter takes place. One interesting difference between a normal

mode and a flutter mode is the following: when the system vibrates in

a normal mode, at one instant of time, the entire energy of the system

will be kinetic; one fourth of the time period later, the energy is com-

pletely potential. But in general, for the system vibrating in flutter

mode, neither the potential nor the kinetic energy completely vanishes at

any instant of time. The mass points in flutter mode vibrate in different

phases and hence to describe the flutter mode graphically we need to show

the deflected configuration at several instances of time within one cycle.

At the end of Chapter II, equations were derived in terms of the

normal coordinates, £ (t), to obtain the response due to externally applied

arbitrary forces. It is now assumed that the flutter mode can be repre-

sented by a series of the first several normal modes with undetermined

coefficients of the form

N
a(y,t) = I ar(y)£r(t)
r=1
(36)
N
w(y,t) = I w (y)£ (t)
r=l

It is further assumed that £ (t) can be obtained by solving the following

N coupled differential equations:

Mr5r(t) + (1 + ig r )<A r £ r (t) = H r (t), r=l,2,3,...,N (37)

 58

where g is the structural damping constant of the r-th mode and

R R
H (t) = - J L(y,t)w (y)dy + / M (y,t)u (y)dy (38)
r r 6a L

In Equation (38), L(y,t) is the lift per unit span and M (y,t) is the
ea

aerodynamic moment per unit span, and these loadings are the result of

the motion of the surface defined by Equation (36). The positive sign

convention used in the unsteady aerodynamic analysis is shown in Figure 13.

Let the motion be simple harmonic with w as the frequency. Then,

a(y,t) = a(y) exp(iu)t)

w(y,t) = w(y)exp(iu)t) (39)

£r(t) = Cr exp(ia)t), r = 1,2,3,...,N

and

L(y,t) = L(y) exp(iu)t)

M (y,t) = M (y) exp(icot) (40)

ea ea

5 (t) = 5 exp(iwt)
r r r

a, w, E, , L(y) , M (y) and H are complex quantities and their phases

are thus defined with respect to some reference vector.

From unsteady aerodynamic theory,

L(y) = L](y)w(y)/b(y) + L2<y)a(y)

(41)
M ea (y) = M1(y)w(y)/b(y) + M2<y)a(y)
 midchord Las t i c a x i s
locat ion

undeflected p o s i t i o n
center of gravity of the blade
Locat i on

•&» X
rt»<*J
wCfrtJ
L <.%D

a.iy) /Uj)
up

2 My)

on
^C
Figure 13. P o s i t i v e Sign Convention for Unsteady Aerodynainic Program.
 60

where

L x (y) Lh(y)/b(y)

L 2 (y)
K^t ,
V y ) " Vy)e(y)]/b(y)
-<
= -irp co b (y)
M]_(y) -M n (y) + L n (y)e(y)
(42)

M 2 (y) H a <y) + [La(y) + M h (y)]e(y)

-Lh(y)£2

and

e(y) = \ + a(y) (43)

The nondimensional aerodynamic coefficients L. , L , M, and M are func-

J
h a ft a

t i o n s of t h e reduced frequency, frequency r a t i o , inflow r a t i o and Mach

number a t t h e spanwise s t a t i o n y and i n case of a m u l t i b l a d e d r o t o r they

a l s o depend on the number of b l a d e s and t h e phase d i f f e r e n c e between the

o s c i l l a t i o n s of t h e b l a d e s . In t h i s t h e s i s , only e q u i v a l e n t s i n g l e bladed

rotor f l u t t e r is considered. This i m p l i e s t h a t for raultibladed rotor sys-

tems, a l l b l a d e s a r e v i b r a t i n g i n some known i n t e r b l a d e mode.

Using Equations ( 3 6 ) , ( 3 8 ) , ( 3 9 ) , ( 4 0 ) , (41) and ( 4 2 ) , i t can now

be shown t h a t

2 A ? -
: = irp u R b ,. ) E A (44)
r *> ref ^ n rn
n=l

where
 61

/• 2 * *
A = JJ [ 6 L , w w +
rn h r n
o
+ 3 (L - L. e)w a * +
a n r n
(45)
+ 3 ^ - L h e)a; w ^ +

+ S^M - (L + M )e + L e 2 } a * a * ]dn
a a h h r n

and

3 = b(n)/bref = 3(n) (46)

and the nondimensional r-th mode shape is defined by

w = w /b _
r r ref
(47)
*
a = a
r r

Now the equation of motion (37) can be written as

r\ ry 9 /

[-oi + ( 1 + i g )oo ]M 5 = TTP (D R b , J £ A (48)

L
°r r r r °° ref n rn
n=l

With the introduction of the nondimensional generalized mass M r defined

b yJ M* = M / ( T T P b 4 _R) o r ,
r r °° r e f

. "in j- 1 —/ N J2.

< - <-^§-> / ^ < +

irp b , o m _
co rer ref (49)

k (n) 2 ,2 / N * .
+ (JL_) 0* . 2(f^)wV}dn
D x. r D _ r r
ref ref
 62

Equation (48) can be written in the completely nondimensional matrix form

as

* ^r 2 *
[-1 + (1 + ig )(—) Z ]M r (U = [A] (O (50)
r co r NJ
(N x N) diagonal (Nxl) (NxN) (Nxl)

Equation (50) is the matrix equation representing a system of N

linear, homogeneous, algebraic equations for the response vector {£}.

For the given angle of incidence at the root and the steady aerodynamic

operating conditions corresponding to a given rotor speed, a structural

dynamic investigation can be carried out which will yield co and M .

Values of g are provided either by estimation or from experiments. If

we then assume a certain frequency of oscillation to, then the unsteady

aerodynamic flow field is completely specified and the matrix [A], can be

evaluated. But since Equation (50) is homogeneous, there arise two cases:

1) if the determinant of the coefficient matrix is nonzero, then { E,} = {0}

is the only solution and this means the system cannot possess a self

excited steady state vibration of this frequency to. 2) if the determinant

of the coefficient matrix is zero, then there exists a nontrivial solu-

tion for {£}, but the amplitude is arbitrary. The lifting surface then

experiences flutter.
 63

CHAPTER IV

THE k-METHOD OF FLUTTER SOLUTION

Statement of the Problem

It was shown in the last chapter that the matrix equation governing

the flutter phenomenon is given by

{ V l + (1 + ig
r
)(w
r
/OJ) 2 ]M*
r^
- [A]HO = {0} (51)

g , a) and M for r = 1 to N are known for a blade if the rotor

r r r

speed, density of the fluid medium and collective pitch are specified.

All the elements of [A] are determined after u) and ft are selected. Since

the terms involved are complicated functions of their arguments, the

flutter solution requires choosing a certain value for ft, scanning through

various appropriate values of to and verifying for each such pair (ft,oo)

whether Equation (51) can be satisfied to yield (ft ,oof ) .

Determinant Method of Solution

This method is based on the fact that for a nontrivial solution of

|^}to exist at flutter., the determinant

A(ftf, Wf) = 0 (52)

where

A(ft, to) = | fi[-l + (1 + ig )(UJ /OJ)2]M - [A] (53)

 64

The coefficient matrix whose determinant is written in Equation

(53) is in general a non-Hermitian matrix and hence A is complex valued.

The procedure begins with choosing a certain value for rotor speed, say

£L . Then a set of scanning frequencies to are used such that the interval

of frequency scanning is small near integral values of frequency ratio.

Then the determinant, A(ft ,u>), is calculated for each case and plotted in

the Argand diagram. If the curve does not pass through the origin, then

ft ^ ftf, and a higher trial rotor speed is chosen and the process repeated.

Figure 14(a) shows one such plot. If the obtained plot is of this

nature, then it is simple to organize a computer program that will inter-

polate for the flutter point (ftf, to ) . Unfortunately, for an example

rotor flutter problem worked out numerically, invariably only a typical

plot as shown in Figure 14(b) was obtained. Such an irregular behavior

of the determinant makes it totally impossible to predict the flutter

point by interpolation and hence requires excessive computer time. The

nature of rotor blade flutter under low inflow conditions seems to be such

that even if (ft ,to) is close to (ftf, oo ) , the determinant value does not

appear to approach the origin. In view of such a numerical difficulty,

this method is abandoned as impracticable.

The Conventional V - g Method or the k Method

This method assumes that for a given rotor speed ft, there exist

several aeroelastic modes. The blade, when disturbed and let free, is

capable of decaying purely in any of these aeroelastic modes, if the ini-

tial conditions are suitable. Each aeroelastic mode is characterized by

(to, g) where to is the damped natural frequency and g is the additional

structural damping required by each of the vacuum normal modes to constitute

Im(A).

-a.C-t-fia)

Figure 14(a) Figure 14(b)

Figure 14. Plot of the F l u t t e r Determinant on t h e Arganet Diagram

 66

this aeroelastic mode. It is proposed from elementary theory that TTg is

equal to the logarithmic decrement of that aeroelastic mode.

The above principle is mathematically modeled below. But since

the subcritical aeroelastic damping cannot be expressed completely

through structural damping, this method cannot predict the damping accu-

rately.

In the flutter equations (51), replace g by (g + g ) . Then approxi-

mate (1 + ig -f ig) by (1 + ig) (1 + ig ) since g and g are small com-

pared to 1 and also the error vanishes as g goes to zero. The flutter

equations now become

([A] +
[<] +
(54)
*(1 + ig r )(l + ig)(o) r /o))V] H U = {0}

If a) =f 0, premultiply Equation (54) by

[(a)
ref / a ) r ) 2 ] / I ( 1 +
^ V

to obtain

s
(w ,/w )
ref r
(i+ig r )M*^
([A] +
w ) - x[i]W U ) = (0}
(55)

where

X = (1 + ig) (o> f /u) (56)

 67

The conventional V - g method used in fixed wing flutter analyses

in the technical literature is modified in this thesis for the rotary

wing analysis. The procedure is as follows:

1. Specify the desired value for collective pitch and density of

the fluid medium.

2. Choose a trial rotor speed ft and conduct a structural dynamic

investigation to yield co and M . Provide g as an estimate or as experi-

r r r
mental data.

3. Choose a suitable set of frequency ratios m.. , m , m ,. . . ,m

as described previously.

4. Corresponding to each of the frequency ratios, do the following

Let the frequency ratio be m. (1 <_ j <_ £). Note that Equation (55) is

in the form of a standard eigenvalue problem. Now each of the N x N com-

plex elements of the eigenvalue matrix (coefficient matrix) can be eval-

uated. The N complex eigenvalues, namely, A-, , A ,...,A can be

obtained. From each complex eigenvalue A , l j ^ i ^ N , w and g are

obtained from Equation (56) as

(j)
to. = co J /Re(A. ( j ) ) (57a)
I ref

(j)
( A a J/
= I (A: )
/ R e ( A UJ/
))/Re(A: )
) (57b)
m i l

since

2
A^j) = (1 + igfj))(w jJ.ih (57c)
I i ref I
 68

A. value for rotor speed i s inferred from

fifj) = ^ j ) / m .
1 1 3

The above procedure is graphically shown below:

A
2 1 ' 81 ' 1

^ ,(D _ (D (1) 0(i)

m —>—

. A (D _+ , ( D pa) 0(D

I . ,(D , CD (D 0 (D

2)
,—- x f - ^ , 8< , «<2)
 69

For convenience, the frequency ratios m 1 _S J _i &» should be

chosen in a monotonically decreasing fashion. For a typical rotor

problem, one list of m could be 7.00, 6.95, 6.90, 6.85,..., 3.00, 2.98,

2.96, 2.94,..., 0.70. For low inflow ratios, scanning should be made

with intervals as low as Am = -0.02, near integral values for the fre-

quency ratio. A case of wake excited flutter has been reported [17] with

frequency ratio as large as 20.0 and if an instability is expected at

high frequencies, suitable values of m must be included in the scanning

list.

Numbering the Aeroelastic Modes

Through the experience gained, it has been observed that when m

is chosen to be high like 7 (with Q = 1) , w , u> u) , . . . ,w

turn out to be close to the vacuum natural frequencies of the N fundamental

modes chosen. The eigenvalues A , A , A„ ,...,A are then

rearranged if necessary, so that the modes can be numbered 1 through N,

closely following the vacuum frequency pattern.

The eigenvalues for m = m , may be disordered compared to the

eigenvalues for m = m . For example, let N = 4, and let A through

A(1) be (200 - 2i), (150 - 1.8i), (132 - 1.651) and (100 - 0.8i). Let

A(2) through A N ( 2 ) be (199 - 1.95i), (131.8 - 1.66i), (100.1 - 0.81i)

and (150.8 - 1.82i). It is clear that the eigenvalues A ' through

(2)
XNV J must be reordered to read (199 - 1.95i), (150.8 - 1.82i), (131.8 -

l*66i) and (100.1 - 0.81i). The new set of eigenvalues A. can be

reordered, (in many cases it will not be necessary), on the basis of

damped natural frequency w. or the damping g. or both.

 70

Desmarais and Bennett [18] have developed a procedure that can be

used to automatically sort the eigenvalues. Basically, they [18] have

introduced a method of determining the eigenvalues of a complex matrix

by iterating on approximate input eigenvalues. They have also discussed

in detail the application of their method to fixed wing flutter problems.

The eigenvalue routine is briefly described in Appendix A. It has been

applied in this thesis to the case of rotor flutter as follows:

1) For m , m and m , any eigenvalue routine like Francis' QR

Transform [19], can be used to obtain the eigenvalues, A. . (These

A. can be further processed through the Desmarais - Bennett subroutine

and refined, if necessary).

2) By inspecting the frequencies obtained above, the eigenvalues

should be ordered, if necessary. The frequencies are generally close

to the undamped natural frequencies.

3) GO. , GO. , GO. , (1 < i < N) , are parabolically extrapolated

1 1 I — —
(4)
to provide an estimate for GO. at m = m. . A similar extrapolated esti-
l 4
(4)
mate is also available for g. . Thus the column of eigenvalues which
(4)
approximate A. is prepared as input to the Desmarais - Bennett sub-
routine along with the eigenvalue matrix for m = m as given by Equation
(55). Then the user selected number of iterations are carried out and

the output eigenvalues are considered to be sufficiently accurate for

(4)
the aeroelastic eigenvalues, A. . Usually five iterations are found

to be sufficient.

4) For obtaining the eigenvalues corresponding to m., first the

elements of the eigenvalue matrix are evaluated (Equation (55)). Then

 71

(j)
the eigenvalues approximate to A are obtained by parabolically extrapo-
lating A. , A. and A. . Then the trial input eigenvalues

are refined by the Desmarais - Bennett subroutine and UK , g. and

1 1
ft. are evaluated as explained previously.

5) Now the frequency - damping - rotor speed plots can be prepared

for all the N aeroelastic modes. The points (ui , ft. ' ) , (w. ,
i i i
(2) (£ ) (£ )
ft. )...,(w. , ft. ) can be j o i n e d i n t h e OJ - ft plane t o y i e l d t h e
p l o t of t h e damped n a t u r a l frequency of the i - t h a e r o e l a s t i c mode a g a i n s t

t h e r o t o r speed, (1 <__ i j< N). By j o i n i n g t h e p o i n t s (g. , ft. ) ,

(2) (2) (£ ) (£ )
(g. , ft.. ) , . . . , ( g . , ft. ) on the g - ft p l a n e , the damping of the
i-th mode is plotted against rotor speed, (1 <_ i <_ N) .

It should be recognized that the rotor speed that is inferred after

the eigenvalue problem is solved, does not necessarily match the rotor

speed that was assumed initially in carrying out the vibrational analysis.

Hence even if one condition for g = 0 was indicated by the analysis, it

may not be a flutter condition since the rotor speed assumed initially may

not match the rotor speed indicated for flutter. The following procedure

is suggested and used in this thesis:

1. Choose a rotor speed ft = ft . Go through the entire procedure,

a.

to obtain the frequency-damping-rotor speed plots.

2. Let the indicated rotor speed for flutter be ft , as deter-

mined by the point corresponding to g = 0. In general, ft ^ ft . Now

D 3.
choose a rotor speed ft = ft, and repeat the analysis for the frequency -
b
damping - rotor speed plots.

3. Continue this procedure until the rotor speed assumed equals

the indicated rotor speed for flutter.

 72

In practice, about 3 or 4 trials will be sufficient for the trial rotor

speed to converge to the flutter speed for unbalanced rotor blades, where

the mass axis is aft of the quarter chord axis. This rotor speed corre-

sponds to a point on the flutter boundary and is called a "matched" flutter

point since compressibility effects in aerodynamics as well as the centrifu-

gal force effects on the structural dynamic properties have been correctly

taken into account. The resulting rotor speed (flutter speed) and the

frequency of oscillation (flutter frequency) are the exact "coupled eigen-

values" of Equation (51). (ft,. = 0, and to = oo are considered "coupled

eigenvalues" here because they together give rise to a nontrivial solution

for {£}, the flutter mode shape vector.)

It is very useful to program the technique in such a way as to obtain

along with the computer output the above mentioned plots, through a

device like the CALCOMP plotter.

The incorporation of the Desmarais-Bennett subroutine makes the

flutter solution automated. When the eigenvalues are obtained in the

fashion mentioned above, a certain amount of computer time is saved com-

pared to the case where only the QR transform is used all the time. But

more importantly, the eigenvalues obtained are almost always ordered in

the desired sequence and hence the frequency - damping - rotor speed plots

can be rapidly prepared. Numerical difficulty and also difficulty in

ordering are expected if any two eigenvalues come close to each other. But

often, when the frequencies become close, the value of damping are well

separated and thus the difficulty does not generally arise. An extensive

computer program was prepared based on this method and a wrong ordering
 73

of the eigenvalues took place only once in about a thousand times. Such

places, in general, can be easily located by an inspection of the computer

plot output.

In the case of fixed wing aeroelastic analysis, the generalized

aerodynamic forces, namely the elements of the [A] matrix, change rela-

tively little from one reduced frequency to another at a given forward

speed. On the other hand, in the case of the low inflow rotor aeroelastic

analysis at a given rotor speed, the elements of the generalized aero-

dynamic matrix generally change very rapidly with frequency ratio, espe-

cially near integral values of frequency ratio. It is now a standard

practice [18] for the fixed wing analysis to compute the [A] matrix for

only a few reduced frequency values, and for intermediate reduced frequen-

cies, the matrix is obtained by interpolation. Because of the reason men-

tioned above, interpolation for the [A] matrix was not resorted to, in

this work.

The Desmarais - Bennett procedure has greatly helped in tracking

the g curves as the frequency ratio is incremented. Thus, in a given

trial, the rotor speed corresponding to g = 0 is automatically computed by

interpolating between the two values of frequency ratio between which g

changes sign. Thus the iteration for the flutter boundary of the rotor

is facilitated, and this is the way an automated matched flutter point

computer program has been developed. This program has been used to solve

several example problems to be discussed later.

A similar procedure for fixed wing flutter can also be formulated.

 74

The flight Mach number and atmospheric density are assumed. The flutter

analysis is carried out for a set of reference reduced frequencies, and

conventional V - g plots are generated to indicate a point (g = 0) on the

flutter boundary. In an incompressible flow, this speed at g = 0 is the

matched flutter point. But in compressible flow, this speed may not be

consistent with the chosen values of density and Mach number for a standard

atmosphere. Now recompute the Mach number according to the indicated

flutter speed and specified density. The entire procedure is repeated until

convergence. In the case of rotor flutter, the dependence of the structural

dynamic properties on the rotor speed has resulted in an additional dimen-

sion of complexity when compared to the fixed wing flutter analysis.

An Example Problem

In Chapter II, the properties of two example blades numbered 1 and 2

were described and their structural dynamic properties were illustrated.

The flutter characteristics of these uniform blades will now be considered.

The k-method as described in the previous section was used to determine the

flutter boundary. The first five normal modes were used in the flutter

analysis. Structural damping coefficients in the vacuum modes were

assumed to be zero.

Hovering flight (V = 0) is assumed for a single bladed model rotor

with

p^ = 0.002378 slug/ft3

a = 1117.0 ft/sec.
00

a (o) = 4 degrees

a = -0.46.
 75

The blade built-in twist is zero. The blade chord is four inches and

uniform along the span. The radius of the rotor is 46.0 inches,

The Mach number at each radial station is given by fty/a . The inflow

ratio variation along the radius is given by the combined simple momentum

and blade element analysis:

2
V —, v / V —( N 2 a ( n ) a (n)na
a
• t \ t °° n)pN . J , °° a(n)as , sv (59)
( ) + / ( } +
ftR " ~ 1 ~ " ftR " ~ 8 ~ 2

where

a = 2bQ/(7TR)

and

a(n) = a ± n c ( n ) / / l - M^(n) (60)

a. (n) i s taken as 5.75. Figure 15 shows the inflow r a t i o as a function

mc
of radius for ft = 90 r a d / s e c .

The frequency r a t i o , m, remains the same for a l l spanwise s t a t i o n s .

But the reduced frequency at any s t a t i o n y i s given by k = mb/y. 130

values of frequency r a t i o were scanned ranging from 12.5 to 0.6 with an

average i n t e r v a l of Am = - 0 . 1 in 7.5 > m > 3.0 and with Am = -0.05 in

3.0 > m > 0.6. The i n t e r v a l was reduced near integer values of m.

The output of the f l u t t e r analysis for Blade No. 1 is summarized

below:
 5.0

4.0 -

3.0

h(?)

2.0 -

1.0 -

0.0
0.0 0.2 0.4 0.6 0.8 1 .0
?
Figure L5. Variation of inflov; Ratio with Blade Radius
 77

Trial Number Trial Rotor Speed Indicated Flutter

Rotor Speed

40 0 rad/sec 100 9 rad/sec

70 5 rad/sec 96 8 rad/sec

96 8 rad/sec 90. 3 rad/sec

90. 3 rad/sec 90. rad/sec

r-i
The flutter boundary rotor speed is concluded to be 90.1 rad/sec.

since convergence is satisfactory. The frequencies and vacuum mode

shapes of this blade at this speed can be approximately seen from Figures

3, and 5(a) through 5(h) of Chapter II.

The flutter frequency is OJ = 195.8 rad/sec. which corresponds to

the frequency ratio of m = 2.17. The flutter mode shape has been deter-

mined to be the eigenvector in Equation (51).

1.0 1.0

-0.8262 + 0.3476 i 0.8963 exp (157.18° i)

-0.2220 + 0.05798 i 0.2294 exp (165.4° i)

:,r
iO = y=A
-0.01212 + 0.007233 i 0.01411 exp (30.85° i)

-0.0002294 + 0.0002815 i 0.0003631 exp (129.2° i)

Figures 16(a), 17(a), 18(a), and 19(a) show the variation of the

damped natural frequencies of the first four aeroelastic modes as

function of rotor speed. Figures 16(b), 17(b), 18(b) and 19(b) show the

variation of the damping present in the first four modes as a function of

rotor speed. These figures will also be used to compare the prediction
 25

k Method

Approximate True
20 Y-g Method

oj.
^
ret
15

10

10 12 16

Figure 16a. F r e q u e n c y - R o t o r Speed PJot ot t h e F i r s t Mode.

-a*
 0.0

•0.1 - A \
A p p r o x i m a t e True
V-g Method
\
•s

\ /
\ / \
\ > \
(}) \
-
I V
\
\
\

\~~~^-~-~^
-0.3 \ ^^--^^
\ ^ ^ ^

\ /
-0.4

\ /
^ ^_y

-0 1 i i 1 1 I 1
4 6 8 10 12 16
X£
Figure 16b. Damping-Rotor Speed P l o t of t h e F i r s t Mode
 50

<v
CO.
m=3-5"
CO
ra

= 2-0
30

20

' k Method
10
Approximate True V-g Method

lu 12 16
jfT
'igure 17a. [• requency-Rotor Speed Plot of the Second Mode
oc
 0.20 k Method

Approximate True V-g Method

(i-)
%

True Flutter Speed

0.0

-0 .10

-0 . 20

16

Figure 17b. D a m p i n g - R o t o r Speed P l o t of t h e Second Mode

 75

(*>
CO.
__5
CO,
rei
rn^ti rr>=7 ^^^

55
m^4-

45

k Method
35
A p p r o x i m a t e T r u e V-g Method

25
10 12 16
JT
igure 18a. Frequencv-Rotor Speed Plot of the. Third Mode
 0.0

(jo
fc

- 0 . 10

•0.15

k Method
-0.20
Approximate True V-g Method

-0.25
10 12 16

Figure 18b. Damping-Rotor Speed Plot of the Third Mode n*

 100

k Method
(j) Approximate True V-g Method

CO,
rei

70

60

50
6 8 10 17 16
SI
Figure 19a Frequency-Rotor Speed Plot of the Fourth Mode
cc
 0.0

-0.02 r ^"XX-'^ ~
% \ "-" ^ ^-" <
y x. \
\ V"^

-0.04

V^'

-0.06 U

" ~" /Approximate True V-g Method

-0.08 k-

-0. 10
_J JL . . _ . . . 1 I | 1 JL
10 12 16
JT
Figure 19b. Damping-Rotor Speed Plot of the Fourth Mode.
 86

of the corresponding quantities by the approximate true V - g method and

the p - k method.

Figures 20(a) and 20(b) illustrate the bending and torsional deforma-

tions of the blade while undergoing flutter at various instants of time.

They were obtained from

N
* * -
wf(n,t) = Re [ I wr(n)?r exp(iwft)]
r=l (61)
N
* *
af(n,t) = Re [ £ a ( n K r exp(iu>ft)]
r=l
Since the modal response vector {£} is composed of vectors making dif-

ferent angles with the real axis in the Argand Diagram, (different rela-

tive phase angles), it follows that when the blade undergoes a flutter

oscillation with constant amplitude some part of the blade is always in

motion. At no instant of time does the kinetic energy or the potential

energy of the system completely vanish. The radial location of the node

point (point at which the instantaneous deflection is zero) continuously

changes with time. The bending and torsional deformations at each radial

station are simple harmonics of a certain amplitude and phase at a fre-

quency equal to the flutter frequency. Since the phase angle at each

radial station has in general a different value, it is necessary to draw

the shape time histories as shown in Figures 20(a) and 20(b) to completely

describe the flutter motion.

The flutter frequency, o)f, is 195.8 rad/sec. at the rotor speed, 0, ,

of 90.1 rad/sec, while the first three vacuum natural frequencies are

90.1, 243.6 and 397.4 rad/sec. The first of these natural frequencies
1.50

0.7

-0.75U

-l.su

0.0 0.2 0.4 0.6 0.8 1.0

9
Figure 20a. Bending Deformation of the Fluttering Blade No. 1
0.4
^ "~" ^—U>t-Tl

«}•»
____——-""""" l~cot= 5K/6
0.2 — —
OOt = I'd/3 —j

r I
1
1
1
-—-^L_ 0.4
0.6 0.8
1.0

cot- 11/2 —/
-0.2 —
— —
- I— cot = n/3

^^^:~:::=:^z^-. i— °°t = o
—
Cot- K/6 —J

03
CO
Figure 20b. Torsional Deformation of the Fluttering Blade No. 1.
 89

corresponds to the flapping mode, the second is predominantly the first

out-of-plane elastic bending mode, and the third is predominantly the

first torsional mode. The flutter mode shape vector consists primarily

of these three modes while the contributions from the fourth and fifth

modes appear almost negligible.

The scales for Figures 20(a) and 20(b) were chosen so as to give the

impression that in this flutter mode, torsional deflection is considerably

larger than bending. It is difficult to compare these two descriptions

of the deformed blade. The maximum amplitude of torsional deflection

occurs approximately at the tip of the blade. The flapping angle (slope

of the bending deflection at the root) is considered the appropriate

quantity in comparing the relative magnitudes of bending and torsion.

The amplitudes of the flutter deformations are arbitrary to a multiplicative

constant, since they are the nontrivial solutions to a homogeneous problem.

The ratio of the maximum amplitude of torsional oscillation to the flap-

ping oscillation is about A.36. It is interesting to note that although

the vacuum torsional frequency is 397.4 rad/sec., this mode is a prime

contributor to flutter which occurs with a frequency as low as 195.8

rad/sec. This emphasizes the importance of including a torsional mode in

a flutter analysis.

The variation of the flutter speed for Blades No. 1 and No. 2 was

studied for various positions of the center of gravity axis. All of the

other dynamic and aerodynamic parameters were left unchanged from the pre-

viously specified values. Figure 21 illustrates the results. This study

confirms the well known result that the flutter speed considerably
 90

midohord
] o c a t ion

r— v

MGO
-05+-
I— e l a s t Lc axis
2-0 *

Cantilevered blade
(Blade No. 2)

hinged blade
(2.23)
(Blade No. 1)
(2.1.7)

(Note: The number written

within brackets indicates
flutter frequency ratio.)

0.0 0.04 0.08 0.L2 0.L6 x 0.24

(-e/ir)
Figure 21. Variation of Flutter Speed with Chordwise
Center of Crnvitv Location.
 91

increases with forward movement of the center of gravity axis. The

flutter speeds (as well as flutter frequencies) of the hinged and fixed

blades are approximately equal.

Because of the large number of factors that enter the flutter

analysis, such a plot as shown in Figure 21 would vary in shape as well

as in magnitude from one rotor system to another. In the case of the

single bladed rotors with Blades No. 1 and 2, the flutter boundary rotor

speed increases with the forward shift of the center of gravity axis.

Experiments conducted by Brooks and Baker [9] showed that forward movement

of the blade chordwise center of gravity location generally raised the

flutter speeds at low pitch angles but had no appreciable effect at high

pitch angles.

An Approximate True V- g Solution

It has been mentioned in the previous sections that when the conven-

tional V- g method is used in a rotor flutter analysis, the indicated

flutter rotor speed does not, in general, match with the chosen trial

rotor speed. This necessitates an iterative procedure so that the indi-

cated flutter speed and trial rotor speed are equal. The frequency -

damping - rotor speed plots of an example problem were shown when the

trial rotor speed (90.27 rad/sec) matched acceptably with the indicated

flutter speed (90.1 rad/sec),

It can be assumed then, that at ^ = 90.1 rad/sec, the second aero-

elastic mode will undergo flutter at u = 195.8 rad/sec. The flutter mode

shape (?) as plotted in Figure 20(a) and 20(b) together with these two

eigenvalues satisfy the flutter equations of motion, expressed by Equation

 92

(51) exactly. This solution is emphasized to be exact because this

solution is totally independent of the method of solution and the motion

is truly simple harmonic with zero damping in this aeroelastic mode.

At all speeds below this speed, every aeroelastic mode possesses

positive damping and hence it can never exhibit simple harmonic motion.

But it is of interest for the blade designer to have some knowledge about

the logarithmic decrement with which each mode will decay at subcritical

rotor speeds. Hence the damping should also be predicted by the flutter

analysis.

The conventional V - g method fails to accomplish this. As is

clear from the analysis, for speeds below the flutter speed, the struc-

tural dynamic description as well as the aerodynamic description are inaccu-

rate because of the lack of matching of the assumed rotor speed and the

rotor speed resulting from the analysis.

If the factor ng is to be used to describe the logarithmic decrement

of the aeroelastic mode, then the applicability of the conventional V - g

method is restricted to a region of rotor speeds very close to the trial

rotor speed. This is the basis for the approximate true V-g method.

A true V - g solution can be considered to be the combination (ft, GO, g)

that exactly satisfies Equation (55). Let the conventional V - g method

be carried out for one trial rotor speed ft,. Whenever for some m. = o)./ft1
1 J J 1
the value ft. J inferred from the eigenvalue A. equals
n ft-,,there is
l I 1
obtained a true V - g solution (ft , OJ . ,g. ) . For one trial rotor speed,
J -^

in general, N such true V - g solutions can be obtained. Thus a large

number of trial rotor speeds can be selected and by obtaining the N such

solutions for each speed, the frequency - damping - rotor speed plots of
 93

the N aeroelastic modes can be generated. This solution is then called

the true V - g solution.

But since the computation required to accomplish the above is exten-

sive, a procedure is described below that provides an approximation to

the true V - g solution.

A certain rotor speed ft is chosen and the conventional V - g method

is employed to generate the frequency - damping - rotor speed plots. A

rotor speed tolerance, Aft , is chosen and only those segments of the

frequency - damping - rotor speed plots lying within ft - Aft g. ft ^ ft + Aft

are considered sufficiently accurate and useful. Then, a higher rotor speed

ft~ is chosen along with Aft and another segment of the plots are generated

for ft - Aft £ ft £ ft + Aft Note ft - Aft is set equal to ft + Aft for

continuity in the results.

In the example flutter analysis of Blade No. 1, the rotor speeds

3)
chosen were 35, 45, 55, 65, 75 and 85 rad/sec. The OJ . , g. plots of

each rotor speed were considered good for Aft = 5 rad/sec. These plot

segments obtained from the rotor speeds are faired to provide a continuous

and smooth result.

Figure 16(a) compares the frequency-rotor speed plot of the first

mode by the k method (conventional V - g method) and the approximate true

"k

V - g method. Aft was chosen to be 0.7. It turns out that the k method

always predicts the frequency of this mode to be approximately the same

as the chosen rotor speed itself. This is quite expected because this

mode is a rigid body flapping mode with vacuum natural frequency being the

rotor speed itself. The prediction in the close neighborhood of the

matched flutter speed is same by both methods.

 94

Figure 16(b) shows the damping in the almost pure first harmonic

rigid body flapping mode.

The erroneous prediction of the frequencies by the k method or

the conventional V - g method at the non-matched rotor speeds, prevents

it from predicting the unsteady aerodynamic damping accurately since the

unsteady aerodynamic forces are highly dependent on the frequency ratio

at low inflow conditions.

In the second aeroelastic mode, the conventional k method predicts

a rapidly oscillating damping while the approximate true V - g method

predicts a more gradual change in damping. For the third mode, both

methods predict similar damping characteristics. This mode being predomi-

nantly torsion, the frequency remains fairly independent of rotor speed

(as predicted by both methods). As the rotor speed increases, the blade

damped natural frequency ratio passes through integer values from about

10 to 4.

It has so far been discussed as to how the state of the art in

employing the k method has been improved for its automated application

in a rotor flutter analysis. The subcritical damping prediction is

almost always unsatisfactory. Hence the conventional k method may be

concluded to be inadequate for subcritical analyses.

The approximate true V - g method, on the other hand, seems to

predict the frequencies and damping well, although comparisons should be

made with experimental results. At any chosen rotor speed, the damped

natural frequencies in the several sought aeroelastic modes can be esti-

mated as follows. If the chosen rotor speed is small, then these fre-

quencies will be close to the undamped natural frequencies. If the

 95

chosen rotor speed is large, then the aerodynamic forces can be expected

to have influenced these frequencies and an approximate estimate must be

made by inspecting the frequencies calculated for lower rotor speeds. The

frequency ratios to be scanned prior to the eigenvalue problem can be

restricted to those values which will give rise to output rotor speeds

which are in the desired range, namely slightly above and below the chosen

rotor speed. In this manner, the computation time can be considerably

reduced and the procedure automated.

 96

CHAPTER V

THE p-k METHOD OF FLUTTER SOLUTION

The Concept of the Decay Rates

It was assumed in Chapter III that the blade motion could be

described through the superposition of normal modes as

N
a(y,t) = I a (y)£ (t)
r=l (62)
w(y,t) = I w (y)4 (t)
r=l

Utilizing this representation, the equations of blade bending and tor-

sional motion are described by

\ i r + U + i g r ) A r ^ = Hj.Ct), r = 1 to N (63)

where

R R
5r(t) = - / L(y,t)wr(y)dy + / M ^ y , t)ar(y)dy (64)
o o

Consider a solution of the form

£r(t) = ^ exp(pt), r = 1 to N (65)

 97

where p = (yoo + iw) and E is the complex amplitude of the r-th mode.

According to this solution, every point on the lifting surface undergoes

a motion of the form

D exp (ywt) cos (wt + tj>)

The amplitude of the oscillatory motion is D exp (ywt), where D is a func-

tion of the y location on the lifting surface. This amplitude decays

exponentially when y < 0. u is the damped frequency of the vibration and

$ is the phase angle with respect to some reference time. It is inter-

esting to note that at every point the motion decays with the same value

of the logarithmic decrement, namely 2TTY .

Since L(y,t) and M (y,t) are linear functions of the normal coordi-
ea
nates, they will also be of the form exp (pt), but possibly with a phase

shift.

N
L(y,t) = L (y)[ J w (y)£ (t)]/b(y) +
J
r=l r (66)
N
+ Vy) (
£ a
r<
y)C
r ( t ) ) = L(y)exp(pt)
r=l

Hence,

_ N __ N
L(y) = L3(y) [ I w r (y)C r ]/b(y) + L4(y) I a r ( y K r (67)
r=l r=l

and, similarly,

N _ N
M (y) = M ( y ) [ £ w ( y ) r ] / b ( y ) + M (y) I a <y)E (68)
J r = 1 r r n r=1 r r
 98

where L , L , M and M are complex unsteady aerodynamic coefficients,

which are functions of p/fi and other parameters. They can be repre-

sented by

L 3 (y) L
hP/b

L 4 (y) [L - L, e]/b
ccp hp
A >- = -TTp 0) b <
(69)
M 3 (y) -M, + L, e
hp hp

M 4 (y) -M + (L + M )e +•
ap ap tip

- ^

where the second subscript, p, denotes that these are p-type aerodynamic

coefficients. An aerodynamic theory that is developed for a motion of

the form exp (pt) is called p-type. A contrasting aerodynamic theory

dealing with simple harmonic motion of the form exp (iwt) is called k-type,

where k is a conventional notation for the reduced frequency associated

with co. The p-k method uses k-type aerodynamics with modifications, to

simulate a p-type motion. The generalized aerodynamic forces of Equation

(64) can be expressed by

H r (t) = E^ exp(pt), r = 1 to N (70)

where

= TTpa) b
2 4 ,-R
2y A £ , r = 1 to N
ref L
. rnp n (71)
n=l
 99

and

2 * *
A = jJf [ 3 L , w w +
rnp hp r n

+ 33(L - L, e) w" a* +
ap hp r n
(72)

3r , * *
+ 3 (Mnp - L,e}
hp ar wn +

+ 3 4 (M - (L + M, )e +
ap ap hp

2, * *
+ L, G } a a JIdn
hp r n

Equation (63) can now be written as

N
P 2 M* £ + ( 1 + i g )w2M* C = w 2 y C A , r=l to N (73)
to L r
r r r r r r n-l, n r nn pp

or in matrix form as

[(p/uO2 + (l+ig r )(w r /o)) 2 ]MM{U = [Ap]{?} (74)

where A is the element on the r-th row and n-th column of matrix [A ]
rnp p
A can be evaluated after obtaining L. , L , M. and M at several
rnp hp ap hp ap
points along the radius by knowing the local Mach number, inflow ratio,

frequency ratio, reduced frequency, decay rate, the number of blades

and the phase angle between blades.

Equation (74) represents the equations of motion that must be

satisfied for the blades to undergo a motion of the form described by

Equation (65). This relation represents N algebraic equations which

are linear, homogeneous, and contain p/co as an unknown quantity.

 100

The overall problem can be stated as follows: For given opera-

ting conditions specified by il, p , a , a (o), etc. is it: possible to

find a value for p, to and {£} such that Equation (74) is satisfied?

If yes, then this solution constitutes an aeroelastic mode. To demon-

strate these aeroelastic modes in a laboratory, the wake must be first

perturbed consistent with the blade vibration pattern.

At one rotor speed, there will be N aeroelastic modes. It is

interesting to note that just like N normal modes, each of these N

aeroelastic modes can be excited independent of each other. Each one is

separately identifiable by a mode shape (£}, a decay rate y and a damped

natural frequency w.

Inference of the Decay Rates from Forced Response

The decay rates of the aeroelastic modes provide a means of

assessing the amount of damping present in the aeroelastic system. It

was pointed out that at each rotor speed, ft, there in general would exist

N aeroelastic modes, each individually characterized by a mode shape, fre-

quency and decay rate. A knowledge of these quantities at sub-flutter

rotor speeds provides a measure of the aeroelastic stability of blade

motion.

Almost all comprehensive unsteady aerodynamic theories are of the

k-type. When a lifting blade surface is acted upon by externally applied

simple harmonic forces, the steady state blade response is also simple

harmonic. The k-type aerodynamic theory will then predict the simple

harmonic airload acting on the blade. From the equations of motion, the

resulting response can be obtained at all points of the surface for any
 101

prescribed simple harmonic forcing function. The response amplitudes

will be functions of the frequency of excitation. As the forcing fre-

quency is varied, the amplitude response can be evaluated as a function

of the forcing frequency, and thus the frequency response function can

be obtained.

In the classical theory of single-degree-of-freedom systems, sev-

eral techniques are available which can be used to predict the decay rate

of the system. These techniques are based on quantities like the frequency

width of the half power points near resonance, the slope of the phase

angle versus frequency at resonance, inference from the in-phase response

and so on. These methods are exact.

For multi-degree-of-freedom systems, the above methods yield approxi-

mate results. If the damped natural frequencies of the various aeroelastic

modes are far apart compared to the frequency width between their half

power points, then the single degree of freedom assumption can be used to

obtain an approximation to the decay rates.

For continuous systems with numerous degrees of freedom, the shape

of the amplitude response curve, namely the plot of the response ampli-

tude versus forcing frequency is dependent on the location of the

response measurement and the point (or points) of application, of the

forcing function. Consequently, for aeroelastic systems, these methods

seem to be somewhat unreliable for predicting the damping. However,

experimentally obtained frequency response characteristics are very use-

ful in substantiating a k-type unsteady aerodynamic theory.. Such a theory

can be used in the p-k method to obtain the decay rates directly.
 102

The Principle of the p-k Method

The p-k method is an approximate technique to obtain the decay rates

by employing a k-type aerodynamic theory in a modified fashion for a p-type

motion. Employing a comprehensive p-type aerodynamic theory is a highly

involved and formidable numerical task in comparison to an equally compre-

hensive k-type aerodynamic theory. The advantages of the p-k method will

be seen in this light. It is a simple, direct, viable and elegant method

for damping prediction which employs k-type aerodynamics. The results are

also used for determining the flutter boundary.

For a lifting surface undergoing motion of the form exp (pt), Equa-

tion (66) describes the instantaneous unsteady lift, as being proportional

to the bending and torsional deflections. In the same equation, the lift

is also written as L(y) exp (pt) which implies that depending on whether

the motion grows or decays, the lift also grows or decays and does so at

the same rate and frequency as that of the motion. It is assumed that the

functions L and L. are not strongly dependent on the magnitude of the

decay rate, at least for slowly converging or diverging motions. This is

the premise of the p - k method.

Mathematically, this assumption then leads to

L, - L, (75)
hp h
L * L
ap a

\ P ~~\

M *M
ap a

for all values of y. Hence the equations of motion given by Equation

(74) will be approximated by

 103

Kp/uO2 + (1 + igr)((Wco)2]M* U> = [A]{?} (76)

^

It can be observed that any error introduced due to the assumption

of Equation (75) vanishes at flutter where indeed y = 0 and k-type aero-

dynamics are valid. Equation (50) of Chapter III was written for the

flutter condition, and it can be verified that Equation (76) is identical

to this relation when y = 0 (or p/w = i ) .

Equation (76) is a system of N linear, homogeneous, algebraic equa-

tions. At any chosen rotor speed, ft, every value of p (if any) is sought

that will produce a non-trivial solution for {£}. In other words, all

possible aeroelastic modes are to be determined.

Substantiation of the p-k Method

Hassig [2] has discussed the p, the k, and the p-k methods of flutter

analysis for fixed wing configurations. The p-k method has been employed

in an actual flutter analysis by Irwin and Guyett [20]. They presented a

graphical method to match the imaginary part of p with the k value of the

aerodynamics.

Hassig [2] has also highlighted, through an example problem, the

differences in frequency and damping prediction by application of the

three methods at various speeds up to flutter. He considered the case of

a twin jet transport aircraft. The flutter equations of motion were first

formulated for the p-type motion using the p-type aerodynamics formulated

by Mazelsky and O'Connell [21].

The damping - frequency - airspeed plots were obtained employing

the p, the k and the p-k methods of solution. The differences in the

damping predicted by the p method and the k method were considerable.

 104

However, there were virtually no differences between the damping pre-

dicted by the p method and the p-k method. In judging the significance

of such an example problem, it should be noted that the unsteady aero-

dynamic theory employed was in a rather simple form. However, it was

sophisticated enough to illustrate the validity of the p-k method.

The p-k method shows considerable promise as a tool for damping

prediction. It is direct, and except for the implied assumption on the

aerodynamic forces, it is an exact method. Furthermore and very import-

antly, the p-k method illustrates the concept of the aeroelastic modes

and their decay rates, by its very formulation. On the contrary, the

conventional k method completely fails in the above regard. However, if

only flutter speed is of interest and not the decay rates at sub-flutter

speeds, then the conventional k method may require significantly less

numerical effort than the p-k method.

Two Numerical Schemes for the p-k Method

The Determinant Iteration Method

Hassig [2] has described the determinant iteration method in

detail. The objective of this method is to determine the value of

p = y<Si + iw, which satisfy Equation (76) at a selected rotor speed. It

is clear that for a nontrivial solution for {£} to exist,

L(p/oo)2 + (1 + igr)(oor/u))2]M* - [A]I = 0 (77)

The above equation can be written symbolically as

Det (£2, p) = 0 (77a)

 105

In the case of a multibladed rotor, the phase differences between the

various blades must be assumed. After choosing the rotor speedft,the

value of p is the only unknown in the elements of the determinant of Equa-

tion (77a). The procedure is as follows:

1. Choose an estimated p.. for the desired aeroelastic mode and

obtain Det (ft, p ) . After choosing another estimated p , obtain

Det (ft, p 2 ) .

2. The Regula Falsi method gives a first iterated value for p as

[p2 Det(ft,P;L) - P;LDet(ft, ?2) ]

3 [Det(ft, p ) ~Det(ft, p )] (78)

3. This process i s repeated according to the recurrence formula

P i+1 Det(ft, Pi ) - p ± Det(ft,p i+1 )

= ( 7 9 )
'i+2 [Det^pJ - Det(ft,pi+1)]

until a specified degree of convergence is obtained. From the converged

value of p, the frequency and damping are given as

w = Im (p) (80)

Y = Re (p)/Im (p)

4. By repeating the above three steps, obtain the converged values

of p for other aeroelastic modes at this same rotor speed ft.

5. Proceed to the next rotor speed and repeat steps 1 through 4.

6. Construct the frequency - damping - rotor speed plots of the

required number of aeroelastic modes up to the desired rotor speed.

 106

In a fixed wing flutter analysis, the generalized aerodynamic

force matrix [A] is normally a relatively smooth function of u). Hence

it is acceptable to calculate [A] at several values of to and utilize its

interpolated result at intermediate values of w. However, in the case

of rotor unsteady aerodynamics, the variation in the aerodynamic coeffi-

cients is large, especially near integral values of equivalent frequency

ratio. While employing the p-k method, this interpolation scheme can be

used to save computation time in the process of convergence.

Evaluation of the complex determinant, Det (£2, p) , is not a particu-

larly difficult task, if N ^ 7. It may be advantageous to first reduce

the matrix to the triangular form.

An Eigenvalue Method of Solution

Equation (76) can be rearranged as

[ 1/M [A]- [(l+igr)(<Wa))2J - ( Y +i) 2 [i] j ( O = (0) (81)

The objective of this method is to determine the pairs (y.u) that will

satisfy Equation (81). It can be observed that Equation (81) is in a

standard eigenvalue form. If a certain value for u) is chosen, then the

matrix

N *•
1/M [A] - (1 + igr)(<Wu))
r

2
can be evaluated and the N complex eigenvalues, (y + i) can be obtained.

Note that for a valid u), the square root of the eigenvalue is (y + i) .

Based on this observation the procedure is as follows:

 107

1. Select an approximation u> = u> for the first aeroelastic

mode. Evaluate the above matrix and obtain all the eigenvalues. In

general, only one of the eigenvalues will be such that its complex

square root will have an imaginary part whose absolute value is close

to 1.

2. That particular root may be expected to lead to a converged

(2)
eigenvalue. A second approximation, u> = to , is chosen and the square

root of the potentially converging eigenvalue is evaluated.

3. An iteration procedure is set up so that for the chosen fre-

2
quency u), the eigenvalue can be expressed as (y + i) . Thus the solution

pair (y, (JO) can be obtained.

4. Repeat steps 1 to 3 for other aeroelastic modes at this rotor

speed ft.

5. Repeat steps 1 to 4 at other rotor speeds and construct the

frequency - rotor speed - decay rate plots of all aeroelastic modes

desired.

It should be noted that once the potentially converging eigenvalue

has been established, the other (N - 1) eigenvalues are of no interest.

Furthermore, every time a new frequency co is chosen in the iteration, an

approximation for the converging eigenvalue is available from the previous

iteration steps. These two observations can be utilized to advantage using

the eigenvalue routine of Desmarais and Bennett [18] as explained in Appen-

dix A.

No comparison has been made regarding the computation time required

by these two approaches for the p-k method, but they are expected to be

comparable. When more than seven degrees of freedom are involved, generally
 108

the eigenvalue approach is better than the determinant approach. In the

example rotor aeroelastic problems to be described, both methods have

been found to be satisfactory. In the second method only w is input and

Y is inferred from the output. In the determinant iteration procedure,

an
both Y d w constitute the input and the iteration is aimed at making both

the real and imaginary parts of the determinant vanish simultaneously.

The eigenvalue method shows promise but more research should be done to

exploit the entire potential of the method.

An Example Problem

In Chapter IV, the damped natural frequencies and decay rates of

example Blade No. 1 were obtained by the conventional k method and the

approximate true V - g method. Now, for the same blade under identical

operating conditions, an aeroelastic analysis is carried out by the p - k

method.

Figure 22(a) shows that the frequency of the first aeroelastic mode

is almost equal to the rotor speed. This should be expected since this is

simply the rigid body flapping mode. Figure 22(b) shows that this mode is

highly damped. The logarithmic decrement of an aeroelastic mode is equal

are
to rrg as well as 2ITY . Hence g and Y related by g = 2Y . Figure 23(a)

illustrates the damped natural frequency of the mode that will exhibit

flutter at high rotor speed. Figure 23(b) shows how the damping rate of

this mode depends on rotor speed. This mode exhibits a hard flutter point

as characterized by the rapidly decreasing damping near the flutter speed.

The flutter mode shape was illustrated in Chapter IV.

The frequency and damping plots of the third aeroelastic mode are

shown in Figures 24(a) and (b). This is a predominantly torsional mode.

 2'j

p-k Method

p_^ Method with h = 80

O O O Approximate True V-g Method

GO
ti)
o. • ^

15

LO

10 12 16
_rr
Figure 22a. I - r e q u e n c y - R o t o r Speed P l o t of t h e F i r s t Mode
o
^c
 0.0

-0.2

%
V
-0.4

-0.6
p-k Method

p-k Method with h - 80

O O O Approximate True V-g Method

-0.8

-1.0
10 12 16
-a"
Figure 22b. Damping-Rotor Speed Plot of the First Mode
o
 50

p-k Method

p-k Method with h = 80

O O O Approximate True V-g Method
coW
COre<
m = 2-5
m- 3 o
30

20

10

10 12 16
_fl
Figure 23a. Frequency-Rotor Speed Plot of the Second Mode.
 p-k Method
0.20
p-k Method, h = 80
W
I2 o o o
/'/
O.li

0.0

-0.10

•0.20-

10 16
Jl'
Figure 23b. Damping-Rotor Speed Plot of the Second Mode
 75

p - k Method w i t h h = 80
65 \-
o 0 O A p p r o x i m a t e True V-g Method

u Q rr\^$ m=7

CO ^
*/ 7 / /"~6 ^"5 rr.~A-

/ ' / ^ /
~C0

45
1"

35 h
1

25 I i 1 I .1 1 I 1
1

10 12 16
H
Figure 24a. Frequency-Rotor Speed Plot of the Third Mode.
 0.0

-0.05

W
I
•0.10

-0.15

— p - k Method
-0.20
p - k Method w i t h h =

O O — O A p p r o x i m a t e True V-g Method

-0.25
10 12 16
JT
Figure 24b. Damping-Rotor Speed Plot of the Third Mode.
 115

At Q, approximately equal to 5.5, the frequency ratio, m equals 9.0, and

as the rotor speed increases, the frequency ratio drops to about 4.0 when

fi equals 12.0. In Figure 24(b), it can be seen that as the frequency

ratio changes by an integer value, the damping curve completes one oscilla-

tion. However, every minimum damping point does not coincide with an

integer frequency ratio. This oscillatory nature of the damping is a

feature of the unsteady rotary wing aerodynamic effects at low inflow con-

ditions. This oscillatory characteristic is due to the close proximity

of the lower wakes and will be discussed in the following section.

The behavior of fourth aeroelastic mode is pictured in Figures

25(a) and (b). This mode is predominantly the second flapwise elastic

bending. The decay rate plot again is oscillatory with integral values

of frequency ratio.

In Figures 22 through 25, the results of the approximate true V - g

method are compared with the p - k method. The comparison is good. It

may be recalled from the previous chapter that for this example problem,

the predictions by the conventional k-method and the approximate true V - g

method were not in good agreement at subcritical speeds. The error as

previously discussed is in the numerical method of solution.

Effect of the Wakes

If the wake spacing is set to infinity in the Jones and Rao [11]

aerodynamic model, the result corresponds to a fixed wing model since

only the bound vorticity and trailing vorticity in the plane of the air-

foil remain. In the computer program, that analyzed the first four aero-

elastic modes of Blade No. 1, by the p - k method, the nondimensional wake

 100 T-

p-k Method

90 -- p-k Method with h = 80

ty o O 0 Approximate True V-g Method

COre
t

m =6 ^.

70 m =/ Jr
^v

/ L^
60 --
„,=<? / o^y /

50 1 1 l I 1 i 1
10 12 16
JTL
Figure 25a. Frequency-Rotor Speed Plot of the Fourth Mode.
 0.0

4-
-0.04

•0.06

p - k Method

•0.08 p - k Method w i t h h = 80

o o o A p p r o x i m a t e T r u e V-g Method

-0.10 1
4 6 8 10 12 16
JT
Figure 25b. Damping-Rotor Speed Plot of the Fourtii Mode
 118

spacing, h, was set to 80.0 at all radial stations. This virtually removes

the effect of the wakes lying below the rotor.

The damped natural frequency of the first mode which is predomi-

nantly the rigid body flapping mode is plotted in Figure 22(a). Figure

22(b) shows that the effect of the lower wakes is to reduce the damping

in this mode.

The frequency of the second aeroelastic mode is well predicted even

after neglecting the lower wakes, as seen in Figure 23(a). However,

Figure 23(b) shows that the damping in this mode does not exhibit the

oscillatory feature. The lower wakes induce aerodynamic pressure forces

on the airfoil and therefore cause the oscillatory damping behavior that

are sensitive to the frequency ratio. The flutter speed is predicted to

be considerably higher in the case where the lower wakes are ignored,

which implies that neglecting them is unconservative. For predicting low

inflow flutter characteristics, the wakes below the rotor plane are

obviously very important.

Figures 24 and 25 show that the frequencies of the third and fourth

aeroelastic modes are estimated well when the lower wakes are ignored.

The amount of damping in these modes do not exhibit an oscillatory nature

as is seen in the case with lower wakes. Interestingly, the damping pre-

dicted is close to the average damping for the case with the lower wakes.

Two other approximate aerodynamic theories have been employed in the

example problem. The first is the steady state aerodynamic theory. The

plunging velocity of the airfoil and the unsteady pitch inclination are

considered to constitute the instantaneous effective angle of attack.

The lift is computed according to steady potential flow theory corresponding

 119

to this angle of attack. The expressions for the aerodynamic coefficients

are

Uh = -2 i/k

L = -2/k 2
a
(82)

M = 0
a

The other simplified aerodynamic theory employed is the quasi

steady-state theory, which neglects the influences of the wake vortex in

the plane of the airfoil. This is equivalent to replacing the Theodorsen

function with the value of 1.0, approached as the reduced frequency goes

to zero, in the expressions for the unsteady lift and moment of the air-

foil about the quarter chord. The aerodynamic coefficients are given by

L = 1 - 2 i/k
h

= I _ _3i _ _2_
a 2 k .2
k
(83)
M^l/2

M = 3/8 - i/k
a

The flutter analyses have been performed by the p - k method employ-

ing these two approximate aerodynamic theories. The decay rates pre-

dicted for the second aeroelastic mode by the two theories are approxi-

mately equal and are shown in Figure 23(b), The flutter is predicted to

occur at a rotor speed of 84.0 radians/second compared to 90.1 radians/

second indicated by the sophisticated rotor unsteady aerodynamic theory

[11]. Though the prediction by the simplified theories is conservative,

 120

it does not reveal the oscillatory nature of the damping. Considering

the fact that the flutter speed prediction by the unsteady fixed-wing

aerodynamic theory is unconservative, it is not possible to relate

definitely as to whether the simplified aerodynamic theories are in

general conservative or not. For accurate analysis, the sophisticated

aerodynamic theories should be resorted to.

A Brief Summary of the Various Methods

Various methods have so far been discussed for the prediction of

flutter and damping characteristics. They are summarized here briefly.

1. The p-method: This is an exact method. The motion is recog-

nized to be an exponentially decaying simple harmonic motion (p-type).

The aerodynamic theory and structural dynamic theory appropriately reflect

the motion.

2. The p-k method: The motion is an exponentially decaying simple

harmonic motion (p-type). The aerodynamic theory suitable for simple

harmonic motion (k-type) is employed after necessary modification to

reflect the p-type motion. The aerodynamic theory may be considered an

approximation, but the method is conceptually sound. The damping values

plotted on the V- g curve are estimates for the actual damping at sub-

critical conditions.

3. The true V- g method: A fictitious structure is considered

which possesses a structural damping of (g + g) in the r-th vacuum normal

mode. At rotor speed fi, this structure can undergo simple harmonic motion

with a frequency w. k-type aerodynamic theory is employed. Since g

varies from one rotor speed to another, and from one aeroelastic mode to
 121

another, the fictitious structure is not a constant entity. We obtain

a matched flutter boundary point for the given structure.

4. Approximate true V - g method: At chosen rotor speeds (at

discrete points) the solution matches with that of the true V - g method.

At other rotor speeds, the solution is an approximation to the true V - g

solution.

5. The k method or the conventional V - g method: k-type aerodynamic

theory is employed. The solution is exact at the matched flutter boundary

point only. The solution obtained at rotor speeds below the matched

flutter speed may not provide any valid information.

 122

CHAPTER VI

UNSTEADY AERODYNAMICS OF THE p-TYPE

The Mathematical Model

In a previous chapter, the various aspects of Loewy's aerodynamic

model were discussed. The model chosen for the p-type aerodynamics is

basically the same except for two differences. The helical vortex con-

tained in the wedge shaped element is not extended to infinity on either

side. Instead, the arc length of the finite wake is simply projected

onto the two dimensional plane of the airfoil without altering the arc

length. The second difference lies in the fact that the wake vorticity

strength is considered to attenuate continuously as the fluid moves in the

downwash of the rotor. These two modifications to the Loewy model [10]

are necessary to analyze the p-type motion of the airfoil. This can be

seen mathematically in the subsequent development.

The concept of an equivalent single bladed rotor has been discussed

previously. In this chapter, an attempt is made to evaluate the differ-

ences between the p-type and p-k type aerodynamic theories for the lift

and moment coefficients of an aerofoil undergoing p-type motion. For

this reason, only a single bladed rotor is considered. Figure 26 shows

the mathematical model with finite lower lying wakes.

Governing Equations

The induced velocity on the airfoil can be obtained by application

of the Biot-Savart law.

 H^st)-
*';§'
t, c ^
V

h<— H r-^(^0
to

£
T7 U*\v
r- ^ j
i*c(<V
r

igure 26. p-type Aerodynamic Mathematical Model for

a Single Bladed Rotor.
 124

bY (F/ft)dC EYoo(C',t)d^
x [ + ! +
V •'> - " ^ * ~iP^TT b "OT^TT-

] (84)
+ I J -™L o—2~T-
n=l -E (x r - V) + n V

In t h e above e q u a t i o n , b , E, n and h f a r e known (n i s the finite

number of lower wakes). The t o t a l c i r c u l a t i o n on the a i r f o i l i s given

by

b
r fc =
^ > / ^ (x' , t ) d x ' (85)
a a
-b

Kutta's wake condition requires that the pressure difference across the

wake (including the trailing edge) vanish, and this mathematically leads

to

(r
V ^ =-yiE l< C » (86)

The vortex strength at any point in the wake is related to the vorticity

at the trailing edge by

Y n o a ' , t ) = Y QO (b,t - At)exp(-uAt), n = 0,1,2,... (87)

where p ^ 0 i s a p r e s c r i b e d v o r t e x v i s c o u s d i s s i p a t i o n f a c t o r , and

At = ( £ ' - b)/V + 2TTii/fi (88)

 125

V= flr' (89)

The kinematic boundary condition must be satisfied on the airfoil

since the fluid velocity relative to the surface of the airfoil can only

be tangential. The vertical induced velocity and the airfoil motion

are related by

"V*''0 = ^ + (x' + I } IE +V*M (90)

where w is the instantaneous downward displacement of the airfoil at the

quarter chord and a(t) is the instantaneous nose up angle of attack of

the flat plate airfoil.

The unsteady Bernoulli equation gives the pressure distribution on

the airfoil as

x
'}
AP(x',t) = p[VY (x',t) + / ~ Yaa',t)dr] (91)
a , aL a
-b

Solution for the Pressure Distribution

Let the airfoil motion be given by

w(t) = w exp (pt) (92)

a(t) = a exp (pt)

Hence

v (xjt) = v (x') exp(pt) (93)

a a

where p is a prescribed complex constant.

 126

When steady state conditions are reached

Y U ' , t) = Y Cx') exp(pt)

a a

Yno(x',t) = Y no (x T ) exp(pt), n = 0,l,2,...,ni (94)

Ap(x',t) = Ap(x') exp(pt)

The following nondimensional quantities are used

x = x'/b

£ = V!h
E = E/b

h = h'/b

co = Im(p) (95)

m = oa/fi

k = oob/V

p = (p + p)/co

P = P/CL)

The Biot-Savart law is simplified to read

-2- - » - - f - » - 7 ^ 1 7 " - v — GOg (96)

where the wake function G(x) is given by

 127

G(x) = G ( x ; k , h , m ; E , u , n ) (97)

E
exp[(l- Ojik]dj
• / (x- O

n
l
+ } exp(-2TT"nm|j)
n=l

E
<" exp[(l - puk](x - Qdg
J ~ 2
2 2
-E (x - £) + n h

From Kutta's condition

^no ( 1 ) X
Ya(C)
-^—--pk/ ^ - d£ (98)

The kinematic boundary condition is written as

v (x) -
-^— = a[l + pk(x +p] + Pk | (99)

For a known airfoil motion, the procedure for solving the pressure

distribution on the airfoil consists of two parts. First, the integral

equation (96) has to be solved for y (x)/V. Then the amplitude of the
cl
pressure distribution is given by the unsteady Bernoulli equation as

ARM _ _!L_ + j k / ^ _ d5 (ioo)

pV -1

The integral equation (96) can be solved by the application of the

Sohngen inversion formula [22]. But in order to gain a physical insight

 128

into the problem the following approach is taken.

The Pressure Mode Approach

Assume that y (x)/V can be written as a combination of various

a
pressure modes:

7 (x) 6 3
— — = A cot ir + I A sin £cf) + £ B cos £<j> (101)
V Z
° 1=1 £=0

where A , A , B are undetermined coefficients, and dp is an airfoil

chordwise coordinate defined by

x = - cos <j> (102)

The cot ((j>/2) term represents the appropriate singular behavior of the

bound vorticity at the leading edge of a flat plate airfoil in a steady

potential flow. The cosine terms would not be present in steady flow.

This is because y and Ap are proportional in steady flow and as Ap van-

a
ishes at the trailing edge, y also does. The unsteady flow relationship
cL

between y and Ap as given by Equation (100) indicates that y does not

a a
necessarily vanish at the trailing edge. In fact, it is proportional to

the time rate of change of bound circulation as indicated by Equation (98).

After substituting Equations (98) and (101) into Equation (96) the

Biot-Savart law becomes

 129

[1 - p k G ( x ) ] A + [-cos <>j - - p k G ( x ) ] A , +
o z 1
+ [-cos 2(J>]A0 + [ - c o s 3<J)]A- + [ - c o s 4*]A. +
Z j 4

2 -
+ [-cos 54>]A_ — [ c o s 6<J>]A, + [-a_ pkG(x)]B
J 0 1 TT O

+ [ a 0 ] B , + [-2xa„ + a. + ~ pkG(x)]B_ +
2 1 Z 1 377 2
v . (x)
+• f ( 4 x - 3)a0 + ~ ] B
J 0 = -2
2 3TT 3 V (103)

where

a = CT/TT, a ? = (2 + xa)/-nr and

a = In [(1 - x ) / ( l + x)] (104)

S i n c e t h e v o r t i c i t y must b e c o n t i n u o u s a t t h e t r a i l i n g e d g e , y (b)/V as

c a l c u l a t e d from E q u a t i o n s (98) and (101) m u s t e q u a l y ( b ) / V a s calculated

3.
from Equation (101). This yields

[-irpkjA + [- -J pk]A_ + [-1 - 2pk]B +

o z l o

+ B x + [| pk - 1]B + B3 = 0 (105)

The eleven undetermined coefficients A , A.,,...,A,, B ,..., B 0 can

o 1 b o 3

be uniquely determined by a collocation method. Ten control points are

chosen on the airfoil where Equation (103) is evaluated. It is desirable

to choose the control points so that they are nearly equidistant from

each other and also not too close to the. leading and trailing edges. This

evaluation of Equation (103) can be written in matrix form as

 130

' [C] {
B } (106)
10 x 11

Equation (105) can be written in a similar fashion as

o =LDJ{;-) (107)

Combining these two matrix equations yields

_V_
I-CI (BJ i - uu$>
B
(108)
LDJ

The unknown pressure mode coefficients can now be obtained from

(B* ) - [El"1 \ ^ V
(109)
0

and the resulting pressure distribution can be computed directly from

Equation (100).

The Lift and Moment Coefficients

If the lift and the moment about the quarter chord of the airfoil

per unit span are represented by L exp (pt) and M exp(pt), then

L = b / Ap(x)dx (110)
-1

M = -b 2 /
r —Ap(x)(x + 1
y)dx (111)
-1
 131

Define a set of non-dimensional coefficients by

— 2 3 — —
L = -Trpco b~ [LH v(w/b) + L a]
hp ap

M = Trpo)2b [M, (w/b) + M a] (112)

np ap

Since this is a linear theory, the pressure distributions due to

pitching and plunging oscillations can be separated to advantage. First

the motion is considered to be plunging and the undetermined coeffi-

cients are evaluated which yield Ln and K. . Then the motion is con-
hp Tip
sidered to be pitching, and L and M are evaluated. The coefficients
ap ap

will be different in each case reflecting the different types of pressure

distributions induced by the two types of airfoil motion. These lift and

moment coefficients are given by

L (or L ) = - - \ [(A + A-/2) +

hp ap 2 o 1

+ pk(3A / 2 + A / 2 + A 2 / 4 ) ] +

l[(B
v - B0/3) +
.2 o 2
rrk
+ pk(B Q + Bx/3 - B 2 /3 - B 3 /5)] ^113)

and
 132

M, (or M ) = (-l/4k2)[(A1 - A_) + pk(4A +

np ap 1 2 o

+ 7A /4 + A 2 /2 - A /4)] +

+ (l/7Tk2)r(-B + 2BJ3 + B 0 /3 +
o 1 2

2B 0 /5) + pk(-5B /3 +
J o

>±/3 + 11B2/15 + B /5)j (114)

Discussion of Results

A FORTRAN Computer program has been prepared to evaluate L , L ,

K. and M as functions of (k, h, m; p, p; E, n j . By setting h = <=°,

hp otp 1

p = i, y = i, E = °°, n = 0 , in this program, the model becomes a fixed

wing two dimensional airfoil undergoing simple harmonic oscillatory motion

in an incompressible flow. The lift and moment coefficients are functions

of the Theodorsen function. The computer program was executed for this

condition over a range of reduced frequency between 0.1 and 0.7, and the

well known results were reproduced. When h is finite, p = i, jj = i, E = °°,

n = °°, the lift and moment coefficients correspond to Loewy's results

[10]. This computer program successfully reproduced these results also,

and hence it is considered that the program has been satisfactorily checked

out.

Three example flow conditions have been chosen as described in the

table below.
 133

Case k h m y/co E n1

1 0.1 3.0 2.3 0.1398 23.0 8

2 0.3 4.5 3.5 0.1262 11.67 8

3 0.5 1.5 6.0 0.1153 12.0 8

It is recalled that u = p + y/w and p == p/u). The airfoil motion is of

the form exp (pcot), and the imaginary part of p is always 1.0. The nega-

tive of the real part of p is defined as the airfoil motion decay factor

and is equal to the logarithmic decrement of the motion divided by 2TT .

For each of the three flow conditions mentioned above, once a

value for the airfoil motion decay factor is chosen, all the parameters

are specified and the aerodynamic coefficients L, , L , M, , M can be

r J
hp ap np otp

evaluated. These coefficients are complex functions because the unsteady

lift and moment are not exactly in phase with the plunging or the pitching

motion of the airfoil. The absolute values of these coefficients are

plotted against the airfoil motion decay factor in Figures 27 through 30.

These coefficients are plotted for each case after being normalized with

respect to the value of that coefficient for simple harmonic motion of

the airfoil, which is the value corresponding to the airfoil motion decay

factor of zero.

For simple harmonic motion, it may be noted that this p-type aero-

dynamic model differs from that of Loewy [10] in two ways. Firstly, the

lengths of the sheets of wake vorticity as well as the number of the

lower sheets of vorticity are finite in the p-type model. Secondly, the
 134

1.2

Case 2 (Loewy)

1. .<d Case 1 (Loewy)

k = 0 . 1, Case 1

L.O

0.9
k= 0 . 5 , Case 3

he!
lU^lo

0.7

Case 3 (Loewy)
.5cr

0.5

t 0.0
.L
0.02
-L
0.04 0.06
±
0.08 0.10

airfoil motion decay factor

ligure 2/. Variation of L, with Airfoil

Motion Decay Factor
 L35

1.2

^ Case 2 (ho
(koewy)
L.i Case 1 (Lo ewy)

k = 0.1, Case

1.0

0.9
k= 0.5, Case 3

1^1
\LJo

0.7

Case 3 (Loewv)
<K"
0.6

0.5

t0.0 0.02 0.04

1
0.06 0.08 0.10

airfoil motion decay factor

Figure 28. Variation of L with Airtoii Motion

Decay Factor.
 136

1.2

1.1
k = 0. 3

k=0.5, Case 3
Case 2 ("Loewy)
1.0
Case 1 (Loewy)

Case 3 (Loewy)
Y k = 0. 1, Case 1

0.9

nhp
M
*'

0.7

0.6

0. 5

1 I I
0.0 0.02 0.04 0.06 0.08 0.10
airfoil motion decay factor

Figure 29. Variation of M with Airfoil Motion

rip
Decay Factor.
 137

k- 0.5, Case 3
Case 2 (Foewy)

Case L (Foewy)
Cases 1,2
(k = 0.1, 0.3)
Case 3 (Foewy)

f
0.0 0.02 0.04 0.06 0.08 0.10

airfoil motion decay factor

Figure 30. Variation ot Ma with Airfoil Motion

r, P
Decay Factor
 138

vortex strength in the wake is allowed to attenuate continuously with

increasing distance downstream. Because of these two differences, for

simple harmonic motion (k type aerodynamics), the values of the coeffi-

cients are different in the two methods. The values from Loewy's theory

are also shown in the figures.

It is observed from the plots that for values of airfoil motion

decay factor up to approximately 0.05, the variation in the coefficients

from their value for simple harmonic motion is less than 5% in all cases.

This substantiates the implied assumption of the p-k method. The differ-

ences between the predictions of the p and the p-k method will depend on

the aeroelastic system being investigated. Figures 27 through 30 compare

only magnitudes of the coefficients, but not their phase. The full

implications of any possible differences in phase can be seen only by a

decay rate analysis.

For an aeroelastic system to be investigated, it would be worth-

while to first evaluate the system characteristics by both the p-k and p

methods, for a typical case. The differences in the results can be con-

sidered to reflect on the accuracy of the p-k method. If for this case

the p-k method is found to be satisfactory, then the remaining cases of

the problem can be analyzed by the p-k method alone.

 139

CHAPTER VII

CONCLUSIONS AND RECOMMENDATIONS

Two relatively new methods of carrying out vibrational analyses

of a nonuniform rotating beam in combined bending and torsion have been

reviewed. The structual dynamic characteristics of an example blade

have been evaluated using the transmission matrix method. The flutter

determinant method, the k method, the approximate true V - g method, and

the p- k method have been employed in an attempt to predict the flutter

speed of an example blade. The principle of the p-k method is explained

from the fundamental concept of decay rates, and an alternative numerical

scheme is proposed for the decay rate solution by this method. An

unsteady rotor aerodynamic theory of the p type has been derived to eval-

uate the implied assumption of the p-k method.

Conclusions

The following conclusions have been drawn from this research pro-

gram:

1. An automated procedure to obtain the matched flutter point

of a rotor blade in an axial flight condition has been developed. A

similar procedure is applicable for determining the matched flutter point

of a fixed wing.

2. The flutter determinant method is not a practicable method for

rotary wing flutter analysis.

 140

3. A method called approximate true V - g method has been devel-

oped. It is illustrated that the errors in the damping prediction at sub-

critical rotor speeds by the k method are largely due to the method of

numerical solution rather than the formulation of the problem.

4. The p- k method is shown to be a viable method for predicting

the damping in several aeroelastic modes at subcritical speeds. An

alternative numerical method of solution to the determinant iteration

procedure of Hassig [2] is provided.

5. Inference of the damping present in the aeroelastic modes from

a frequency response plot for external simple harmonic excitation is not

a reliable procedure.

6. In rotary wing flutter analyses, the vorticity lying in the

downwash of the rotor should not be neglected in the unsteady aerodynamic

theory, because the ignoring of this vorticity results in an unconserva-

tive flutter speed estimation.

7. An unsteady rotor aerodynamic theory of the p type has been

developed. The variation of the unsteady lift and moment coefficients

with respect to the airfoil motion decay factor indicates that the

implied assumption of the p - k method is sound.

8. The p - k method shows considerable promise and may become a

standard method of the future.

Recommendations

The following suggestions are made regarding future research in

the area of rotor aeroelasticity.

 141

1. Since the p - k method has considerable potential, more

research may be carried out regarding application of this method to

the various areas of rotor aeroelasticity including the case of a

helicopter rotor in forward flight.

2. Further work may be done to improve the eigenvalue method

of solution to the p - k method.

3. More p type unsteady aerodynamic formulations may be devel-

oped so that the validity of the p - k method can be established for a

variety of aeroelastic systems.

4. Since comprehensive experimental results are invaluable in

substantiating any analytical model, more experiments may be planned to

verify the predictions of the p - k method.

 142

APPENDIX A

THE EIGENVALUE ROUTINE OF DESMARAIS AND BENNETT

A general
Iterated
complex matrix Desmarais and eigenvalues,{A.}
[C], N x N Bennett J
->-
Eigenvalue (Presumably accurate
Approximate
Routine eigenvalues of [C])
eigenvalues of
[C], {A.}
A
Number of iterations
specified by the user

a) The matrix [C] is transformed to upper Hessenberg form [H] by Gaussian

elimination.

b) The trial eigenvalues are improved by Laguerre iteration as follows:

The first approximate eigenvalue is iterated the specified number of

times. Then the second approximate eigenvalue is iterated and so on until

the N-th eigenvalue is iterated the specified number of times. With this

the program ends.

The iteration on the j-th eigenvalue consists of the following

steps:

1. The trial eigenvalue = A.

J
2. The quantity

f(T) = |[H] - T[I] (A-l)

and its first two derivatives f (T) and f'Cx) are computed at x = A
 143

from Hyman's recurrence relations [23].

3. The improved eigenvalue A. is given by

ij - A . - n £ / { S l [ l + / ( n r l ) ( - l +V ySl2) ] (A-2)

where

I = j -1 (A-3)

n£ = N - £ (A-4)

f'(A.)
S
l =
f(A.)
" 1=1I(A. - A.)
(A-5)
3 J i

- x-, 2
f'(A.) fM(A.) 1
1
S
2 = _ J_ _ (A-6)
f(A.) f(X.) y
L
1
J 3 — 2
1=1 (A. - A.)
J i

Note that A. is the i-th eigenvalue of [C], and i < j. The terms with Z

sign must be ignored if j = 1.

4. Set A. = A.. Repeat steps 2 and 3 above until the number

J J
of iterations has been completed.

5. After the specified number of iterations have been carried

out, set A. = A., the iterated value at the end of the last iteration.
J J
Thus the j-th eigenvalue is obtained-
 REFERENCES

1. Pierce, G. A. and white, W. F,, "Unsteady Rotor Aerodynamics at

Low Inflow and Its Effect on Flutter," AIAA Paper No. 72-959,
presented at 2nd AIAA Atmospheric Flight Mechanics Conference,
Palo Alto, California (September 11-13, 19 72).

2. Hassig, H. J., "An Approximate True Damping Solution of the

Flutter Equation by Determinant Iteration," J. of Aircraft, Vol.
8, No. 11, November 1971, pp. 885-889.

3. Gessow, A., and Myers, Jr., G. C , Aerodynamics of the Helicopter,

Frederick Ungar Publishing Co., New York, N. Y., 1967.

4. Houbolt, J. C. and Brooks, G. W., "Differential Equations of Motion

for Combined Flapwise Bending, Chordwise Bending, and Torsion of
Twisted Nonuniform Rotor Blades," NACA Technical Report 1346, 1958.

5. Nagaraja, K. S. S., "Analytical and Experimental Aeroelastic Studies

of a Helicopter Rotor in Vertical Flight," Ph.D. Thesis, Georgia
Institute of Technology, 1975.

6. Murthy, V. R., "Determination of the Structural Dynamic Character-

istics of Rotor Blades and the Effect of Phase Angle on Multibladed
Rotor Flutter," Ph.D. Thesis, Georgia Institute of Technology, 1975.

7. Rubin, S., "Review of Mechanical Immittance and Transmission

Matrix Concepts," J. of Acoustical Society of America, Vol. 41,
No. 5, May 1967, pp. 1171-1179.

8. Hunter, W. F., "Integrating Matrix Method for Determining the

Natural Vibration Characteristics of Propeller Blades," NASA
Technical Note D-6064, 1970.

9. Brooks, G. N. and Baker, J. E., "An Experimental Investigation

of the Effect of Various Parameters Including the Tip Mach Number
on the Flutter of Some Helicopter Rotor Models," NACA Technical
Note 4005, 1958.

10. Loewy, R. G., "A Two Dimensional Approximation to the Unsteady

Aerodynamics of Rotary Wings," Journal of the Aerospace Sciences,
Vol. 24, No. 2, pp. 81-92, 144, February 1957.

11. Jones, W. P. and Rao, B. M., "Compressibility Effects on Oscilla-

ting Rotor Blades in Hovering Flight," AIAA Journal, Vol. 8, No.
2, pp. 321-329, February 19 70.
 145

12. Jones, W. P., "The Oscillating Airfoil in Subsonic Flow,"

British Aeronautical Research Council, R & M 2921, 1956.

13. Hammond, C. E. and Pierce, G. A., "A Compressible Unsteady

Aerodynamic Theory for Helicopter Rotors," presented at the
AGARD Specialists' Meeting on "The Aerodynamics of Rotary Wings,"
Marseille, France, September 13-15, 1972.

14. White, W. F., Jr., "Effect of Compressibility on Three Dimensional

Helicopter Rotor Blade Flutter," Ph.D. Thesis, Georgia Institute
of Technology, School of Aerospace Engineering, August 1972.

15. Ham, N. D., Moser, H. H. and Zvara, J., "Investigation of Rotor

Response to Vibrating Aerodynamic Inputs, Part I. Experimental
Results and Correlation with Theory," U. S. Air Force, Air Research
and Development Command, WADC TR 58-87, AD 203389, October 1958.

16. Daughaday, H., Du Walt, F. and Gates, C , "Investigation of Heli-

copter Blade Flutter and Load Amplification Problems," Journal of
the American Helicopter Society, Vol. 2, No. 3, pp. 27-45, July
1957.

17. Theodorsen, T. and Regier, A. A., "Effect of the Lift Coefficient

on Propeller Flutter," NACA ACR L5F30, July 1954.

18. Desmarais, R. N., and Bennett, R. M., "An Automated Procedure for
Computing Flutter Eigenvalues," Journal of Aircraft, Volume 11,
No. 2, Feb. 1974, pp. 75-80.

19. Francis, J. G. F., "The QR Transformation," Parts I and II,

Computer Journal, Vol. 4, Oct. 1961, Jan. 1962, pp. 265-271 and
232-245.

20. Irwin, C. A. K. and Guyett, P. R., "The Subcritical Response and

Fluttej of a Swept Wing Model," Tech. Rept. 65186, Aug. 1965, Royal
Aircraft Establishment, Farnborough, U. K., also, Aeronautical
Research Council R & M 349 7, London, England.

21. Mazelsky, B. and O'Connell, R. F., "Transient Aerodynamic Proper-

ties of Wings: Review and Suggested Electrical Representation for
Analog Computers," LR 11577, July 1956, Lockheed Aircraft Corp.,
California Div., Burbank, Calif.

22. Sohngen, H., Die Losungen der Integralgleichung und deren Anwendung
in der Tragfl'ugeltheorie, Math. Z., Band 45, pp. 245-264, 1939.

23. Wilkinson, J. H., The Algebraic Eigenvalue Problem, Clarendon

Press, Oxford, 1967, pp. 426-427.
 146

VITA

Sathy Padmanaban Viswanathan was born to Janaki Jambunathan and

Sathy Gopalan Padmanaban on April 5, 1949 in Coimbatore, Tamil Nadu,

India. After attending the T.A.R. Chettiar High School and Suburban

High School, he joined A. M. Jain College in Madras. He studied

Bachelor of Technology course in Aeronautical Engineering in the Indian

Institute of Technology at Madras from 1965 to 1970, and obtained first

rank in the University. He received M.S. Degree in Aerospace Engineering

from Georgia Institute of Technology, Atlanta, Ga. in 19 72. Presently he

is employed by Bell Helicopter Textron in Fort Worth, Texas as a Senior

Dynamics Engineer in Rotor Dynamics Group.